diff --git a/docs/source/api_reference/embodichain/embodichain.toolkits.rst b/docs/source/api_reference/embodichain/embodichain.toolkits.rst index cc2639d5..d773b9f9 100644 --- a/docs/source/api_reference/embodichain/embodichain.toolkits.rst +++ b/docs/source/api_reference/embodichain/embodichain.toolkits.rst @@ -1,4 +1,4 @@ -embodichain.toolkits +embodichain.toolkits ==================== .. automodule:: embodichain.toolkits @@ -11,12 +11,88 @@ urdf_assembly -GraspKit --------- +GraspKit — Parallel-Gripper Grasp Sampling +------------------------------------------- -.. automodule:: embodichain.toolkits.graspkit +The ``embodichain.toolkits.graspkit.pg_grasp`` module provides a complete pipeline for generating antipodal grasp poses for parallel-jaw grippers. The pipeline consists of three stages: + +1. **Antipodal sampling** — Surface points are uniformly sampled on the mesh and rays are cast to find antipodal point pairs on opposite sides. +2. **Pose construction** — For each antipodal pair, a 6-DoF grasp frame is built aligned with the approach direction. +3. **Filtering & ranking** — Grasp candidates that cause the gripper to collide with the object are discarded; survivors are scored by a weighted cost. + +.. rubric:: Public API + +.. currentmodule:: embodichain.toolkits.graspkit.pg_grasp + +The main entry point is :class:`GraspGenerator`. It is configured via :class:`GraspGeneratorCfg` and :class:`GripperCollisionCfg`. + +.. autosummary:: + :nosignatures: + + GraspGenerator + GraspGeneratorCfg + AntipodalSampler + AntipodalSamplerCfg + GripperCollisionChecker + GripperCollisionCfg + ConvexCollisionChecker + ConvexCollisionCheckerCfg + + +GraspGenerator +~~~~~~~~~~~~~~~ + +.. autoclass:: GraspGenerator + :members: generate, annotate, get_grasp_poses, visualize_grasp_pose + :show-inheritance: + +GraspGeneratorCfg +~~~~~~~~~~~~~~~~~~ + +.. autoclass:: GraspGeneratorCfg + :members: + :show-inheritance: + +AntipodalSampler +~~~~~~~~~~~~~~~~~ + +.. autoclass:: AntipodalSampler + :members: sample + :show-inheritance: + +AntipodalSamplerCfg +~~~~~~~~~~~~~~~~~~~~ + +.. autoclass:: AntipodalSamplerCfg + :members: + :show-inheritance: + +GripperCollisionChecker +~~~~~~~~~~~~~~~~~~~~~~~~ + +.. autoclass:: GripperCollisionChecker + :members: query + :show-inheritance: + +GripperCollisionCfg +~~~~~~~~~~~~~~~~~~~~ + +.. autoclass:: GripperCollisionCfg + :members: + :show-inheritance: + +ConvexCollisionChecker +~~~~~~~~~~~~~~~~~~~~~~~ + +.. autoclass:: ConvexCollisionChecker + :members: query, query_batch + :show-inheritance: + +ConvexCollisionCheckerCfg +~~~~~~~~~~~~~~~~~~~~~~~~~~ + +.. autoclass:: ConvexCollisionCheckerCfg :members: - :undoc-members: :show-inheritance: diff --git a/docs/source/features/toolkits/grasp_generator.rst b/docs/source/features/toolkits/grasp_generator.rst new file mode 100644 index 00000000..ba77e77b --- /dev/null +++ b/docs/source/features/toolkits/grasp_generator.rst @@ -0,0 +1,242 @@ +Generating and Executing Robot Grasps +====================================== + +.. currentmodule:: embodichain.lab.sim + +This tutorial demonstrates how to generate antipodal grasp poses for a target object and execute a full grasp trajectory with a robot arm. It covers scene initialization, robot and object creation, interactive grasp region annotation, grasp pose computation, and trajectory execution in the simulation loop. + +The Code +~~~~~~~~ + +The tutorial corresponds to the ``grasp_generator.py`` script in the ``scripts/tutorials/grasp`` directory. + +.. dropdown:: Code for grasp_generator.py + :icon: code + + .. literalinclude:: ../../../../scripts/tutorials/grasp/grasp_generator.py + :language: python + :linenos: + + +The Code Explained +~~~~~~~~~~~~~~~~~~ + +Configuring the simulation +-------------------------- + +Command-line arguments are parsed with ``argparse`` to select the number of parallel environments, the compute device, and optional rendering features such as ray tracing and headless mode. + +.. literalinclude:: ../../../../scripts/tutorials/grasp/grasp_generator.py + :language: python + :start-at: def parse_arguments(): + :end-at: return parser.parse_args() + +The parsed arguments are passed to ``initialize_simulation``, which builds a :class:`SimulationManagerCfg` and creates the :class:`SimulationManager` instance. When ray tracing is enabled a directional :class:`cfg.LightCfg` is also added to the scene. + +.. literalinclude:: ../../../../scripts/tutorials/grasp/grasp_generator.py + :language: python + :start-at: def initialize_simulation(args) -> SimulationManager: + :end-at: return sim + +Creating a robot and a target object +------------------------------------ + +A UR10 arm with a parallel-jaw gripper is created via :meth:`SimulationManager.add_robot`. The gripper URDF and drive properties are configured so that the arm joints and finger joints can be controlled independently. + +.. literalinclude:: ../../../../scripts/tutorials/grasp/grasp_generator.py + :language: python + :start-at: def create_robot(sim: SimulationManager + :end-at: return sim.add_robot(cfg=cfg) + +The target object (a mug) is loaded as a :class:`objects.RigidObject` from a PLY mesh file: + +.. literalinclude:: ../../../../scripts/tutorials/grasp/grasp_generator.py + :language: python + :start-at: def create_mug(sim: SimulationManager): + :end-at: return mug + +Annotating and computing grasp poses +------------------------------------- + +Grasp generation is performed by :class:`~embodichain.toolkits.graspkit.pg_grasp.GraspGenerator`, which runs an antipodal sampler on the object mesh. The mesh data (vertices and triangles) is extracted from the :class:`objects.RigidObject` via its accessor methods. A :class:`~embodichain.toolkits.graspkit.pg_grasp.GraspGeneratorCfg` controls sampler parameters (sample count, gripper jaw limits) and the interactive annotation workflow: + +1. Open the visualization in a browser at the reported port (e.g. ``http://localhost:11801``). +2. Use *Rect Select Region* to highlight the area of the object that should be grasped. +3. Click *Confirm Selection* to finalize the region. + +After annotation, antipodal point pairs are cached to disk and automatically reused unless user call `GraspGenerator.annotate()`. + +For each environment, a grasp pose is computed by calling :meth:`~embodichain.toolkits.graspkit.pg_grasp.GraspGenerator.get_grasp_poses` with the object pose and desired approach direction. The result is a ``(4, 4)`` homogeneous transformation matrix representing the grasp frame in world coordinates. Set ``visualize=True`` to open an Open3D window showing the selected grasp on the object. + +The approach direction is the unit vector along which the gripper approaches the object. In this tutorial, we use a fixed approach direction (straight down in world frame) for simplicity, but it can be customized based on the task or object geometry. + +.. literalinclude:: ../../../../scripts/tutorials/grasp/grasp_generator.py + :language: python + :start-at: gripper_collision_cfg = GripperCollisionCfg( + :end-at: logger.log_info(f"Get grasp pose cost time: {cost_time:.2f} seconds") + +Building and executing the grasp trajectory +------------------------------------------- + +Once a grasp pose is obtained, a waypoint trajectory is built that moves the arm from its rest configuration to an approach pose (offset above the grasp), down to the grasp pose, closes the fingers, lifts, and returns. The trajectory is interpolated for smooth motion and executed step-by-step in the simulation loop. + +.. literalinclude:: ../../../../scripts/tutorials/grasp/grasp_generator.py + :language: python + :start-at: def get_grasp_traj(sim: SimulationManager + :end-at: return interp_trajectory + +Configuring GraspGeneratorCfg +------------------------------ + +:class:`~embodichain.toolkits.graspkit.pg_grasp.GraspGeneratorCfg` controls the overall grasp annotation workflow. The key parameters are listed below. + +.. list-table:: GraspGeneratorCfg parameters + :header-rows: 1 + :widths: 25 15 60 + + * - Parameter + - Default + - Description + * - ``viser_port`` + - ``15531`` + - Port used by the Viser browser-based visualizer for interactive grasp region annotation. + * - ``use_largest_connected_component`` + - ``False`` + - When ``True``, only the largest connected component of the object mesh is used for sampling. Useful for meshes that contain disconnected fragments. + * - ``antipodal_sampler_cfg`` + - ``AntipodalSamplerCfg()`` + - Nested configuration for the antipodal point sampler. See the table below for its parameters. + * - ``max_deviation_angle`` + - ``π / 12`` + - Maximum allowed angle (in radians) between the specified approach direction and the axis connecting an antipodal point pair. Pairs that deviate more than this threshold are discarded. + +The ``antipodal_sampler_cfg`` field accepts an :class:`~embodichain.toolkits.graspkit.pg_grasp.AntipodalSamplerCfg` instance, which controls how antipodal point pairs are sampled on the mesh surface. + +.. list-table:: AntipodalSamplerCfg parameters + :header-rows: 1 + :widths: 25 15 60 + + * - Parameter + - Default + - Description + * - ``n_sample`` + - ``20000`` + - Number of surface points uniformly sampled from the mesh before ray casting. Higher values yield denser coverage but increase computation time. + * - ``max_angle`` + - ``π / 12`` + - Maximum angle (in radians) used to randomly perturb the ray direction away from the inward normal. Larger values increase diversity of sampled antipodal pairs. Setting this to ``0`` disables perturbation and samples strictly along surface normals. + * - ``max_length`` + - ``0.1`` + - Maximum allowed distance (in metres) between an antipodal pair. Pairs farther apart than this value are discarded; set this to match the maximum gripper jaw opening width. + * - ``min_length`` + - ``0.001`` + - Minimum allowed distance (in metres) between an antipodal pair. Pairs closer together than this value are discarded to avoid degenerate or self-intersecting grasps. + +Configuring GripperCollisionCfg +------------------------------- + +:class:`~embodichain.toolkits.graspkit.pg_grasp.GripperCollisionCfg` models the geometry of a parallel-jaw gripper as a point cloud and is used to filter out grasp candidates that would collide with the object. All length parameters are in metres. + +.. list-table:: GripperCollisionCfg parameters + :header-rows: 1 + :widths: 25 15 60 + + * - Parameter + - Default + - Description + * - ``max_open_length`` + - ``0.1`` + - Maximum finger separation of the gripper when fully open. Should match the physical gripper specification. + * - ``finger_length`` + - ``0.08`` + - Length of each finger along the Z-axis (depth direction from the root). Should match the physical gripper specification. + * - ``y_thickness`` + - ``0.03`` + - Thickness of the gripper body and fingers along the Y-axis (perpendicular to the opening direction). + * - ``x_thickness`` + - ``0.01`` + - Thickness of each finger along the X-axis (parallel to the opening direction). + * - ``root_z_width`` + - ``0.08`` + - Extent of the gripper root block along the Z-axis. + * - ``point_sample_dense`` + - ``0.01`` + - Approximate number of sample points per unit length along each edge of the gripper point cloud. Higher values produce denser point clouds and improve collision-check accuracy at the cost of additional computation. + * - ``max_decomposition_hulls`` + - ``16`` + - Maximum number of convex hulls used when decomposing the object mesh for collision checking. More hulls give a tighter shape approximation but increase cost. + * - ``open_check_margin`` + - ``0.01`` + - Extra clearance added to the gripper open length during collision checking to account for pose uncertainty or mesh inaccuracies. + + +The Code Execution +~~~~~~~~~~~~~~~~~~ + +To run the script, execute the following command from the project root: + +.. code-block:: bash + + python scripts/tutorials/grasp/grasp_generator.py + +A simulation window will open showing the robot and the mug. A browser-based visualizer will also launch (default port ``11801``) for interactive grasp region annotation. + +You can customize the run with additional arguments: + +.. code-block:: bash + + python scripts/tutorials/grasp/grasp_generator.py --num_envs --device --enable_rt --headless + +After confirming the grasp region in the browser, the script will compute a grasp pose, print the elapsed time, and then wait for you to press **Enter** before executing the full grasp trajectory in the simulation. Press **Enter** again to exit once the motion is complete. + + +Grasp Annotation CLI +~~~~~~~~~~~~~~~~~~~~ + +EmbodiChain provides a dedicated CLI for interactively annotating grasp regions on a mesh and caching the resulting antipodal point pairs, without requiring a full simulation environment. + +Basic usage:: + + python -m embodichain annotate-grasp --mesh_path /path/to/object.ply + +This will: + +1. Load the mesh file via ``trimesh``. +2. Launch a browser-based annotator (default port ``15531``). +3. Open http://localhost:15531 in your browser, use *Rect Select Region* to highlight the graspable area, then click *Confirm Selection*. +4. Compute antipodal point pairs on the selected region and cache them to disk. + +Common options:: + + python -m embodichain annotate-grasp \ + --mesh_path /path/to/object.ply \ + --viser_port 15531 \ + --n_sample 20000 \ + --max_length 0.1 \ + --min_length 0.001 \ + +.. list-table:: CLI options + :header-rows: 1 + :widths: 25 15 60 + + * - Option + - Default + - Description + * - ``--mesh_path`` + - *(required)* + - Path to the mesh file (``.ply``, ``.obj``, ``.stl``, etc.). + * - ``--viser_port`` + - ``15531`` + - Port for the browser-based annotation UI. + * - ``--n_sample`` + - ``20000`` + - Number of surface points to sample for antipodal pair detection. + * - ``--max_length`` + - ``0.1`` + - Maximum distance (metres) between antipodal pairs; should match the gripper's maximum opening width. + * - ``--min_length`` + - ``0.001`` + - Minimum distance (metres) between antipodal pairs; filters out degenerate pairs. + * - ``--device`` + - ``cpu`` + - Compute device (``cpu`` or ``cuda``). diff --git a/docs/source/features/toolkits/index.rst b/docs/source/features/toolkits/index.rst index b6f0b22d..f6886746 100644 --- a/docs/source/features/toolkits/index.rst +++ b/docs/source/features/toolkits/index.rst @@ -7,4 +7,5 @@ ToolKits convex_decomposition urdf_assembly + grasp_generator \ No newline at end of file diff --git a/embodichain/__main__.py b/embodichain/__main__.py index 5125296b..522ca48f 100644 --- a/embodichain/__main__.py +++ b/embodichain/__main__.py @@ -20,6 +20,7 @@ python -m embodichain preview-asset --asset_path /path/to/asset.usda --preview python -m embodichain run-env --env_name my_env + python -m embodichain annotate-grasp --mesh_path /path/to/object.ply """ from __future__ import annotations @@ -63,6 +64,17 @@ def main() -> None: run_env_parser.set_defaults(func=run_env_cli) + # -- annotate-grasp ------------------------------------------------------ + annotate_grasp_parser = subparsers.add_parser( + "annotate-grasp", + help="Interactively annotate grasp region on a mesh.", + ) + from embodichain.toolkits.graspkit.scripts.annotate_grasp import ( + cli as annotate_grasp_cli, + ) + + annotate_grasp_parser.set_defaults(func=annotate_grasp_cli) + # -- Parse --------------------------------------------------------------- # If no sub-command is given, print help and exit. if len(sys.argv) < 2 or sys.argv[1] in ("-h", "--help"): diff --git a/embodichain/lab/sim/objects/rigid_object.py b/embodichain/lab/sim/objects/rigid_object.py index 0c8477e2..24de293b 100644 --- a/embodichain/lab/sim/objects/rigid_object.py +++ b/embodichain/lab/sim/objects/rigid_object.py @@ -949,25 +949,51 @@ def set_body_type(self, body_type: str) -> None: self.body_type = body_type - def get_vertices(self, env_ids: Sequence[int] | None = None) -> torch.Tensor: + def get_vertices( + self, env_ids: Sequence[int] | None = None, scale: bool = False + ) -> torch.Tensor: """ Retrieve the vertices of the rigid objects. Args: env_ids (Sequence[int] | None): A sequence of environment IDs for which to retrieve vertices. If None, retrieves vertices for all instances. + scale (bool): Whether to multiply the vertices by the body scale. Defaults to False. Returns: torch.Tensor: A tensor containing the user IDs of the specified rigid objects with shape (N, num_verts, 3). """ ids = env_ids if env_ids is not None else range(self.num_instances) - return torch.as_tensor( + verts = torch.as_tensor( np.array( [self._entities[id].get_vertices() for id in ids], ), dtype=torch.float32, device=self.device, ) + if scale: + verts = verts * self.get_body_scale(env_ids).unsqueeze_(1) + return verts + + def get_triangles(self, env_ids: Sequence[int] | None = None) -> torch.Tensor: + """ + Retrieve the triangle indices of the rigid objects. + + Args: + env_ids (Sequence[int] | None): A sequence of environment IDs for which to retrieve triangle indices. + If None, retrieves triangle indices for all instances. + + Returns: + torch.Tensor: A tensor containing the triangle indices of the specified rigid objects with shape (N, num_tris, 3). + """ + ids = env_ids if env_ids is not None else range(self.num_instances) + return torch.as_tensor( + np.array( + [self._entities[id].get_triangles() for id in ids], + ), + dtype=torch.int32, + device=self.device, + ) def get_user_ids(self, env_ids: Sequence[int] | None = None) -> torch.Tensor: """Get the user ids of the rigid bodies. diff --git a/embodichain/toolkits/graspkit/pg_grasp/__init__.py b/embodichain/toolkits/graspkit/pg_grasp/__init__.py index 82c25ce0..d9719a08 100644 --- a/embodichain/toolkits/graspkit/pg_grasp/__init__.py +++ b/embodichain/toolkits/graspkit/pg_grasp/__init__.py @@ -14,8 +14,7 @@ # limitations under the License. # ---------------------------------------------------------------------------- -from .antipodal import AntipodalGenerator, GraspSelectMethod - -__all__ = ["AntipodalGenerator", "GraspSelectMethod"] - -del antipodal +from .antipodal_sampler import * +from .collision_checker import * +from .gripper_collision_checker import * +from .antipodal_generator import * diff --git a/embodichain/toolkits/graspkit/pg_grasp/antipodal.py b/embodichain/toolkits/graspkit/pg_grasp/antipodal.py deleted file mode 100644 index 1b7b7f1a..00000000 --- a/embodichain/toolkits/graspkit/pg_grasp/antipodal.py +++ /dev/null @@ -1,670 +0,0 @@ -# ---------------------------------------------------------------------------- -# Copyright (c) 2021-2026 DexForce Technology Co., Ltd. -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. -# ---------------------------------------------------------------------------- - -import open3d as o3d -import numpy as np -import time -import pathlib -import pickle -import os - -from enum import Enum -from copy import deepcopy -from typing import List - -from .cone_sampler import ConeSampler -from embodichain.utils.utility import get_mesh_md5 -from embodichain.utils import logger - - -class GraspSelectMethod(Enum): - NORMAL_SCORE = 0 - NEAR_APPROACH = 1 - CENTER = 2 - - -class AntipodalGrasp: - def __init__(self, pose: np.ndarray, open_len: float, score: float) -> None: - self.pose = pose # [4, 4] of float grasp pose - self.open_len = open_len # gripper open length - self.score = score # grasp pose score - - def grasp_pose_visual_mesh(self, gripper_open_length: float = None): - if gripper_open_length is None: - gripper_open_length = self.open_len - open_ratio = 1.0 - else: - open_ratio = self.open_len / gripper_open_length - open_ratio = max(1e-4, open_ratio) - gripper_finger = o3d.geometry.TriangleMesh.create_box( - gripper_open_length * 0.04, - gripper_open_length * 0.04, - gripper_open_length * 0.5, - ) - gripper_finger.translate( - np.array( - [ - -gripper_open_length * 0.02, - -gripper_open_length * 0.02, - -gripper_open_length * 0.25, - ] - ) - ) - gripper_left = deepcopy(gripper_finger) - gripper_left = gripper_left.translate( - np.array( - [ - -gripper_open_length * open_ratio * 0.5, - 0, - -gripper_open_length * 0.25, - ] - ) - ) - - gripper_right = deepcopy(gripper_finger) - gripper_right = gripper_right.translate( - np.array( - [gripper_open_length * open_ratio * 0.5, 0, -gripper_open_length * 0.25] - ) - ) - - gripper_root1 = o3d.geometry.TriangleMesh.create_box( - gripper_open_length * open_ratio, - gripper_open_length * 0.04, - gripper_open_length * 0.04, - ) - gripper_root1.translate( - np.array( - [ - gripper_open_length * open_ratio * -0.5, - gripper_open_length * -0.02, - gripper_open_length * -0.02, - ] - ) - ) - gripper_root1.translate( - np.array( - [ - 0, - 0, - gripper_open_length * -0.5, - ] - ) - ) - - gripper_root2 = o3d.geometry.TriangleMesh.create_box( - gripper_open_length * 0.04, - gripper_open_length * 0.04, - gripper_open_length * 0.5, - ) - gripper_root2.translate( - np.array( - [ - gripper_open_length * -0.02, - gripper_open_length * -0.02, - gripper_open_length * -0.25, - ] - ) - ) - gripper_root2.translate( - np.array( - [ - 0, - 0, - gripper_open_length * -0.75, - ] - ) - ) - - gripper_visual = gripper_left + gripper_right + gripper_root1 + gripper_root2 - gripper_visual.compute_vertex_normals() - return gripper_visual - - -class Antipodal: - def __init__( - self, - point_a: np.ndarray, - point_b: np.ndarray, - normal_a: np.ndarray, - normal_b: np.ndarray, - ) -> None: - """antipodal contact pair - - Args: - point_a (np.ndarray): position in point a - point_b (np.ndarray): position in point b - normal_a (np.ndarray): normal in point a - normal_b (np.ndarray): normal in point b - """ - self.point_a = point_a - self.point_b = point_b - self.normal_a = normal_a - self.normal_b = normal_b - self.dis = np.linalg.norm(point_a - point_b) - self.angle_cos = self.normal_a.dot(-self.normal_b) - self.score = self._get_score() - self._canonical_pose = self._get_canonical_pose() - - def _get_score(self): - # TODO: only normal angle is taken into account - return self.angle_cos - - def get_dis(self, another) -> float: - """get distance acoording to another antipodal - - Args: - other (Antipodal): another antipodal - - Returns: - float: distance - """ - aa_dis = np.linalg.norm(self.point_a - another.point_a) - bb_dis = np.linalg.norm(self.point_b - another.point_b) - ab_dis = np.linalg.norm(self.point_a - another.point_b) - ba_dis = np.linalg.norm(self.point_b - another.point_a) - return min(aa_dis, bb_dis, ab_dis, ba_dis) - - def get_dis_arr(self, others) -> np.ndarray: - """get distance acoording to others antipodals - - Args: - others (List[Antipodal]): other antipodals - - Returns: - np.ndarray: distance array - """ - if not others: - return np.array([], dtype=float) - # Vectorized extraction of points using list comprehension and np.array - other_a = np.array([o.point_a for o in others], dtype=float) - other_b = np.array([o.point_b for o in others], dtype=float) - aa_dis = np.linalg.norm(other_a - self.point_a, axis=1) - ab_dis = np.linalg.norm(other_a - self.point_b, axis=1) - ba_dis = np.linalg.norm(other_b - self.point_a, axis=1) - bb_dis = np.linalg.norm(other_b - self.point_b, axis=1) - dis_arr = np.vstack([aa_dis, ab_dis, ba_dis, bb_dis]).min(axis=0) - return dis_arr - - def _get_canonical_pose(self) -> np.ndarray: - """get canonical pose of antipodal contact pair - - Returns: - np.ndarray: canonical pose - """ - # assume gripper closing along x_axis - x_d = self.point_a - self.point_b - x_d = x_d / np.linalg.norm(x_d) - y_d = np.cross(np.array([0, 0, 1.0], dtype=float), x_d) - if np.linalg.norm(y_d) < 1e-4: - y_d = np.cross(np.array([1, 0, 0.0], dtype=float), x_d) - y_d = y_d / np.linalg.norm(y_d) - z_d = np.cross(x_d, y_d) - pose = np.eye(4, dtype=float) - pose[:3, 0] = x_d # rotation x - pose[:3, 1] = y_d # rotation y - pose[:3, 2] = z_d # rotation z - pose[:3, 3] = (self.point_a + self.point_b) / 2 # position - return pose - - def sample_pose(self, sample_num: int = 36) -> np.ndarray: - """sample parallel gripper grasp poses given antipodal contact pairs - - Args: - sample_num (int, optional): sample number. Defaults to 36. - - Returns: - np.ndarray: [sample_num, 4, 4] of float. Sample poses. - """ - # assume gripper closing along x_axis - x_d = self._canonical_pose[:3, 0] - y_d = self._canonical_pose[:3, 1] - z_d = self._canonical_pose[:3, 2] - position = self._canonical_pose[:3, 3] - beta_list = np.linspace(2 * np.pi / sample_num, 2 * np.pi, sample_num) - grasp_poses = np.empty(shape=(sample_num, 4, 4), dtype=float) - for i in range(sample_num): - sample_z = np.sin(beta_list[i]) * y_d + np.cos(beta_list[i]) * z_d - sample_z = sample_z / np.linalg.norm(sample_z) - sample_y = np.cross(sample_z, x_d) - pose = np.eye(4, dtype=float) - pose[:3, 0] = x_d # rotation x - pose[:3, 1] = sample_y # rotation y - pose[:3, 2] = sample_z # rotation z - pose[:3, 3] = position # position - grasp_poses[i] = pose - return grasp_poses - - -class AntipodalGenerator: - def __init__( - self, - mesh_o3dt: o3d.t.geometry.TriangleMesh, - open_length: float, - min_open_length: float = 0.002, - max_angle: float = np.pi / 10, - surface_sample_num: int = 5000, - layer_num: int = 4, - sample_each_layer: int = 6, - nms_ratio: float = 0.02, - antipodal_sample_num: int = 16, - unique_id: str = None, - cache_dir: str = None, - ): - """antipodal grasp pose generator - - Args: - mesh_o3dt (o3d.t.geometry.TriangleMesh): input mesh - open_length (float): gripper maximum open length - max_angle (float, optional): maximum grasp direction with surface normal. Defaults to np.pi/10. - surface_sample_num (int, optional): contact sample number in mesh surface. Defaults to 5000. - layer_num (int, optional): cone sample layer number . Defaults to 4. - sample_each_layer (int, optional): cone sample number in each layer. Defaults to 6. - nms_ratio (float, optional): nms distance ratio. Defaults to 0.02. - antipodal_sample_num (int, optional): grasp poses sample on each antipodal contact pair. Defaults to 16. - cache_dir (str, optional): file cache directory. Defaults to None. - """ - self._antipodal_max_angle = max_angle - self._open_length = open_length - self._min_open_length = min_open_length - self._mesh_o3dt = mesh_o3dt - verts = mesh_o3dt.vertex.positions.numpy() - self._center_of_mass = verts.mean(axis=0) - if unique_id is None: - unique_file_name = self._get_unique_id( - mesh_o3dt, open_length, max_angle, surface_sample_num - ) - else: - unique_file_name = f"{unique_id}_{str(open_length)}_{str(max_angle)}_{str(surface_sample_num)}" - if cache_dir is None: - cache_dir = os.path.join(pathlib.Path.home(), "grasp_pose") - logger.log_debug(f"Set cache directory to {cache_dir}.") - if not os.path.isdir(cache_dir): - os.mkdir(cache_dir) - cache_file = os.path.join(cache_dir, f"{unique_file_name}.pickle") - if not os.path.isfile(cache_file): - # generate cache - grasp_list = self._generate_cache( - cache_file, - mesh_o3dt=mesh_o3dt, - max_angle=max_angle, - surface_sample_num=surface_sample_num, - layer_num=layer_num, - sample_each_layer=sample_each_layer, - nms_ratio=nms_ratio, - antipodal_sample_num=antipodal_sample_num, - ) - else: - # load cache - grasp_list = self._load_cache(cache_file) - self._grasp_list = grasp_list - - def _get_unique_id( - self, - mesh_o3dt: o3d.t.geometry.TriangleMesh, - open_length: float, - max_angle: float, - surface_sample_num: int, - ) -> str: - mesh_md5 = get_mesh_md5(mesh_o3dt) - return ( - f"{mesh_md5}_{str(open_length)}_{str(max_angle)}_{str(surface_sample_num)}" - ) - - def _generate_cache( - self, - cache_file: str, - mesh_o3dt: o3d.t.geometry.TriangleMesh, - max_angle: float = np.pi / 10, - surface_sample_num: int = 5000, - layer_num: int = 4, - sample_each_layer: int = 6, - nms_ratio: float = 0.02, - antipodal_sample_num: int = 16, - ): - self._mesh_o3dt = mesh_o3dt - self._cone_sampler = ConeSampler( - max_angle=max_angle, - layer_num=layer_num, - sample_each_layer=sample_each_layer, - ) - mesh_o3dt = mesh_o3dt.compute_vertex_normals() - assert 1e-4 < max_angle < np.pi / 2 - self._mesh_len = self._get_pc_size(mesh_o3dt.vertex.positions.numpy()).max() - start_time = time.time() - antipodal_list = self._antipodal_generate(mesh_o3dt, surface_sample_num) - logger.log_debug( - f"Antipodal sampling cost {(time.time() - start_time) * 1000} ms." - ) - logger.log_debug(f"Find {len(antipodal_list)} initial antipodal pairs.") - - valid_antipodal_list = self._antipodal_clean(antipodal_list) - - start_time = time.time() - nms_antipodal_list = self._antipodal_nms( - valid_antipodal_list, nms_ratio=nms_ratio - ) - logger.log_debug(f"NMS cost {(time.time() - start_time) * 1000} ms.") - logger.log_debug( - f"There are {len(nms_antipodal_list)} antipodal pair after nms." - ) - # all poses - start_time = time.time() - grasp_poses, score, open_length = self._sample_grasp_pose( - nms_antipodal_list, antipodal_sample_num - ) - logger.log_debug(f"Pose sampling cost {(time.time() - start_time) * 1000} ms.") - logger.log_debug( - f"There are {grasp_poses.shape[0]} poses after grasp pose sampling." - ) - # write data - data_dict = { - "grasp_poses": grasp_poses, - "score": score, - "open_length": open_length, - } - pickle.dump(data_dict, open(cache_file, "wb")) - # TODO: contact pair visualization - # self.antipodal_visual(nms_antipodal_list) - grasp_num = grasp_poses.shape[0] - logger.log_debug(f"Write {grasp_num} poses to pickle file {cache_file}.") - # Use list comprehension for efficient list construction - grasp_list = [ - AntipodalGrasp(grasp_poses[i], open_length[i], score[i]) - for i in range(grasp_num) - ] - return grasp_list - - def _load_cache(self, cache_file: str): - data_dict = pickle.load(open(cache_file, "rb")) - grasp_num = data_dict["grasp_poses"].shape[0] - logger.log_debug(f"Load {grasp_num} poses from pickle file {cache_file}.") - # Use list comprehension for efficient list construction - grasp_list = [ - AntipodalGrasp( - data_dict["grasp_poses"][i], - data_dict["open_length"][i], - data_dict["score"][i], - ) - for i in range(grasp_num) - ] - return grasp_list - - def _get_pc_size(self, vertices): - return np.array( - [ - vertices[:, 0].max() - vertices[:, 0].min(), - vertices[:, 1].max() - vertices[:, 1].min(), - vertices[:, 2].max() - vertices[:, 2].min(), - ] - ) - - def _antipodal_generate( - self, mesh_o3dt: o3d.t.geometry.TriangleMesh, surface_sample_num: int = 5000 - ): - surface_pcd = mesh_o3dt.to_legacy().sample_points_uniformly(surface_sample_num) - points = np.array(surface_pcd.points) - normals = np.array(surface_pcd.normals) - point_num = points.shape[0] - scene = o3d.t.geometry.RaycastingScene() - scene.add_triangles(mesh_o3dt) - # raycast - ray_n = self._cone_sampler._ray_num - ray_num = point_num * ray_n - ray_begins = np.empty(shape=(ray_num, 3), dtype=float) - ray_direcs = np.empty(shape=(ray_num, 3), dtype=float) - origin_normals = np.empty(shape=(ray_num, 3), dtype=float) - origin_points = np.empty(shape=(ray_num, 3), dtype=float) - start_time = time.time() - for i in range(point_num): - ray_direc = self._cone_sampler.cone_sample_direc( - normals[i], is_visual=False - ) - # raycast from outside of object - ray_begin = points[i] - 2 * self._mesh_len * ray_direc - ray_direcs[i * ray_n : (i + 1) * ray_n, :] = ray_direc - ray_begins[i * ray_n : (i + 1) * ray_n, :] = ray_begin - origin_normals[i * ray_n : (i + 1) * ray_n, :] = normals[i] - origin_points[i * ray_n : (i + 1) * ray_n, :] = points[i] - logger.log_debug(f"Cone sampling cost {(time.time() - start_time) * 1000} ms.") - - start_time = time.time() - rays = o3d.core.Tensor( - np.hstack([ray_begins, ray_direcs]), dtype=o3d.core.Dtype.Float32 - ) - logger.log_debug(f"Open3d raycast {(time.time() - start_time) * 1000} ms.") - - raycast_rtn = scene.cast_rays(rays) - hit_dis_all = raycast_rtn["t_hit"].numpy() - hit_normal_all = raycast_rtn["primitive_normals"].numpy() - - # max_angle_cos = np.cos(self._antipodal_max_angle) - antipodal_list = [] - # get antipodal points - start_time = time.time() - for i in range(ray_num): - hit_dis = hit_dis_all[i] - hit_normal = hit_normal_all[i] - hit_point = ray_begins[i] + ray_direcs[i] * hit_dis - antipodal_dis = np.linalg.norm(origin_points[i] - hit_point) - if ( - antipodal_dis > self._min_open_length - and antipodal_dis < self._open_length - ): - # reject thin close object - antipodal = Antipodal( - origin_points[i], hit_point, origin_normals[i], hit_normal - ) - antipodal_list.append(antipodal) - logger.log_debug( - f"Antipodal initialize cost {(time.time() - start_time) * 1000} ms." - ) - return antipodal_list - - def _sample_grasp_pose( - self, antipodal_list: List[Antipodal], antipodal_sample_num: int = 36 - ): - antipodal_num = len(antipodal_list) - grasp_poses = np.empty( - shape=(antipodal_sample_num * antipodal_num, 4, 4), dtype=float - ) - scores = np.empty(shape=(antipodal_sample_num * antipodal_num,), dtype=float) - open_length = np.empty( - shape=(antipodal_sample_num * antipodal_num,), dtype=float - ) - for i in range(antipodal_num): - grasp_poses[i * antipodal_sample_num : (i + 1) * antipodal_sample_num] = ( - antipodal_list[i].sample_pose(antipodal_sample_num) - ) - scores[i * antipodal_sample_num : (i + 1) * antipodal_sample_num] = ( - antipodal_list[i].score - ) - open_length[i * antipodal_sample_num : (i + 1) * antipodal_sample_num] = ( - antipodal_list[i].dis - ) - return grasp_poses, scores, open_length - - def get_all_grasp(self) -> List[AntipodalGrasp]: - """get all grasp - - Returns: - List[AntipodalGrasp]: list of grasp - """ - return self._grasp_list - - def select_grasp( - self, - approach_direction: np.ndarray, - select_num: int = 20, - max_angle: float = np.pi / 10, - select_method: GraspSelectMethod = GraspSelectMethod.NORMAL_SCORE, - ) -> List[AntipodalGrasp]: - """Select grasps. Masked by max_angle and sort by grasp score - - Args: - approach_direction (np.ndarray): gripper approach direction - select_num (int, optional): select grasp number. Defaults to 10. - max_angle (float, optional): max angle threshold (angle with surface normal). Defaults to np.pi/10. - select_method (select_method, optional) - Return: - List[AntipodalGrasp]: list of grasp - """ - grasp_num = len(self._grasp_list) - all_idx = np.arange(grasp_num) - - # Vectorized extraction of poses and scores using list comprehension - grasp_poses = np.array([g.pose for g in self._grasp_list], dtype=float) - scores = np.array([g.score for g in self._grasp_list], dtype=float) - position = grasp_poses[:, :3, 3] - - # mask acoording to table up direction - grasp_z = grasp_poses[:, :3, 2] - direc_dot = (grasp_z * approach_direction).sum(axis=1) - valid_mask = direc_dot > np.cos(max_angle) - valid_id = all_idx[valid_mask] - - # sort acoording to different grasp score - if select_method == GraspSelectMethod.NORMAL_SCORE: - valid_score = scores[valid_id] - sort_valid_idx = np.argsort(valid_score)[::-1] # large is better - elif select_method == GraspSelectMethod.NEAR_APPROACH: - position_dot = (position * approach_direction).sum(axis=1) - valid_height = position_dot[valid_id] - sort_valid_idx = np.argsort(valid_height) - elif select_method == GraspSelectMethod.CENTER: - center_dis = np.linalg.norm(position - self._center_of_mass, axis=1) - valid_center_dis = center_dis[valid_id] - sort_valid_idx = np.argsort(valid_center_dis) - else: - logger.log_warning(f"select_method {select_method.name} not implemented.") - # return all grasp - return self._grasp_list - - # get best score sample index - result_num = min(len(sort_valid_idx), select_num) - best_valid_idx = sort_valid_idx[:result_num] - best_idx = valid_id[best_valid_idx] - - # Use list comprehension for faster list construction - return [self._grasp_list[idx] for idx in best_idx] - - def _antipodal_nms( - self, antipodal_list: List[Antipodal], nms_ratio: float = 0.02 - ) -> List[Antipodal]: - antipodal_num = len(antipodal_list) - nms_mask = np.empty(shape=(antipodal_num,), dtype=bool) - nms_mask.fill(True) - score_list = np.empty(shape=(antipodal_num,), dtype=float) - - for i in range(antipodal_num): - score_list[i] = antipodal_list[i].score - - sort_idx = np.argsort(score_list)[::-1] - - dis_th = self._mesh_len * nms_ratio - for i in range(antipodal_num): - if not nms_mask[sort_idx[i]]: - continue - antipodal_max = antipodal_list[sort_idx[i]] - other_antipodal = [] - other_idx = [] - for j in range(i + 1, antipodal_num): - if nms_mask[sort_idx[j]]: - other_antipodal.append(antipodal_list[sort_idx[j]]) - other_idx.append(sort_idx[j]) - dis_arr = antipodal_max.get_dis_arr(other_antipodal) - invalid_mask = dis_arr < dis_th - for j, flag in enumerate(invalid_mask): - if flag: - nms_mask[other_idx[j]] = False - nms_antipodal_list = [] - for i in range(antipodal_num): - if nms_mask[sort_idx[i]]: - nms_antipodal_list.append(antipodal_list[sort_idx[i]]) - - # TODO: nms validation check. remove in future - # antipodal_num = len(nms_antipodal_list) - # for i in range(antipodal_num): - # for j in range(i + 1, antipodal_num): - # antipodal_dis = nms_antipodal_list[i].get_dis(nms_antipodal_list[j]) - # if antipodal_dis < dis_th: - # logger.log_warning(f"find near antipodal {i} and {j} with dis {antipodal_dis}") - return nms_antipodal_list - - def _antipodal_clean(self, antipodal_list: List[Antipodal]): - # TODO: need collision checker - - valid_antipodal = [] - max_angle_cos = np.cos(self._antipodal_max_angle) - for antipodal in antipodal_list: - if ( - 1e-5 < antipodal.dis < self._open_length - and antipodal.angle_cos > max_angle_cos - ): - valid_antipodal.append(antipodal) - return valid_antipodal - - def antipodal_visual(self, antipodal_list): - mesh_visual = self._mesh_o3dt.to_legacy() - antipodal_num = len(antipodal_list) - draw_points = np.empty(shape=(antipodal_num * 2, 3), dtype=float) - draw_lines = np.empty(shape=(antipodal_num, 2), dtype=int) - for i in range(antipodal_num): - direc = antipodal_list[i].point_b - antipodal_list[i].point_a - direc = direc / np.linalg.norm(direc) - anti_begin = antipodal_list[i].point_a - direc * 0.005 - anti_end = antipodal_list[i].point_b + direc * 0.005 - draw_points[i * 2] = anti_begin - draw_points[i * 2 + 1] = anti_end - draw_lines[i] = np.array([i * 2, i * 2 + 1], dtype=int) - - line_set = o3d.geometry.LineSet( - points=o3d.utility.Vector3dVector(draw_points), - lines=o3d.utility.Vector2iVector(draw_lines), - ) - draw_colors = np.empty(shape=(antipodal_num, 3), dtype=float) - draw_colors[:] = np.array([0.0, 1.0, 1.0]) - line_set.colors = o3d.utility.Vector3dVector(draw_colors) - o3d.visualization.draw_geometries([line_set, mesh_visual]) - - def grasp_pose_visual( - self, grasp_list: List[AntipodalGrasp] - ) -> List[o3d.t.geometry.TriangleMesh]: - """visualize grasp pose - - Args: - grasp_list (List[AntipodalGrasp]): list of grasp - - Returns: - List[o3d.t.geometry.TriangleMesh]: list of visualization mesh - """ - pose_num = len(grasp_list) - visual_mesh_list = [self._mesh_o3dt.compute_vertex_normals()] - - max_angle_cos = np.cos(self._antipodal_max_angle) - - for i in range(pose_num): - grasp_mesh = grasp_list[i].grasp_pose_visual_mesh( - gripper_open_length=self._open_length - ) - grasp_mesh.transform(grasp_list[i].pose) - # low score: red | hight score: blue - score_ratio = (grasp_list[i].score - max_angle_cos) / (1 - max_angle_cos) - score_ratio = min(1.0, score_ratio) - score_ratio = max(0.0, score_ratio) - grasp_mesh.paint_uniform_color(np.array([1 - score_ratio, 0, score_ratio])) - visual_mesh_list.append(o3d.t.geometry.TriangleMesh.from_legacy(grasp_mesh)) - return visual_mesh_list diff --git a/embodichain/toolkits/graspkit/pg_grasp/antipodal_generator.py b/embodichain/toolkits/graspkit/pg_grasp/antipodal_generator.py new file mode 100644 index 00000000..f6389ff8 --- /dev/null +++ b/embodichain/toolkits/graspkit/pg_grasp/antipodal_generator.py @@ -0,0 +1,768 @@ +# ---------------------------------------------------------------------------- +# Copyright (c) 2021-2026 DexForce Technology Co., Ltd. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# ---------------------------------------------------------------------------- + +import os +import argparse +import open3d as o3d +import time +import torch +import numpy as np +import trimesh +import hashlib +import torch.nn.functional as F + +import viser +import viser.transforms as tf +from pathlib import Path +from typing import Any, cast + +from embodichain.utils import logger +from embodichain.utils import configclass +from embodichain.toolkits.graspkit.pg_grasp.antipodal_sampler import ( + AntipodalSampler, + AntipodalSamplerCfg, +) +from embodichain.utils import configclass +from embodichain.toolkits.graspkit.pg_grasp import ( + GripperCollisionChecker, + GripperCollisionCfg, +) + +GRASP_ANNOTATOR_CACHE_DIR = ( + Path.home() / ".cache" / "embodichain" / "grasp_annotator_cache" +) +GRASP_ANNOTATOR_CACHE_DIR.mkdir(parents=True, exist_ok=True) + + +__all__ = ["GraspGenerator", "GraspGeneratorCfg"] + + +@configclass +class GraspGeneratorCfg: + """Configuration for :class:`GraspGenerator`. + + Controls the interactive grasp region annotation workflow, including the + browser-based visualizer settings, antipodal sampling parameters, and + grasp-pose filtering thresholds. + """ + + viser_port: int = 15531 + """Port used by the Viser browser-based visualizer for interactive grasp + region annotation.""" + + use_largest_connected_component: bool = False + """When ``True``, only the largest connected component of the selected mesh + region is retained. Useful for meshes that contain disconnected fragments + or when selecting a local feature such as a handle.""" + + antipodal_sampler_cfg: AntipodalSamplerCfg = AntipodalSamplerCfg() + """Nested configuration for the antipodal point sampler. Controls the + number of sampled surface points, ray perturbation angle, and gripper jaw + distance limits. See :class:`AntipodalSamplerCfg` for details.""" + + max_deviation_angle: float = np.pi / 12 + """Maximum allowed angle (in radians) between the specified approach + direction and the axis connecting an antipodal point pair. Pairs that + deviate more than this threshold from perpendicular to the approach are + discarded during grasp pose computation.""" + + +class GraspGenerator: + """Antipodal grasp-pose generator for parallel-jaw grippers. + + Given an object mesh, ``GraspGenerator`` produces feasible grasp poses + through a three-stage pipeline: + + 1. **Antipodal sampling** — Surface points are uniformly sampled and + rays are cast along (and near) the inward normal to find antipodal + point pairs on opposite sides of the mesh (:meth:`generate`). + Alternatively, an interactive Viser-based annotator lets a human + select the graspable region (:meth:`annotate`). + 2. **Pose construction** — For each antipodal pair, a 6-DoF grasp + frame is built so that the gripper opening aligns with the pair axis + and the approach direction is consistent with a user-specified + vector (:meth:`get_grasp_poses`). + 3. **Filtering & ranking** — Grasp candidates that would cause the + gripper to collide with the object are discarded. Surviving poses + are scored by a weighted cost that penalises angular deviation from + the approach direction, narrow opening length, and distance to the + mesh centroid. + + Typical usage:: + + generator = GraspGenerator(vertices, triangles, cfg=cfg) + + # Programmatic: sample on the whole mesh or a sub-region + generator.generate() # whole mesh + generator.generate(face_indices=some_idx) # specific faces + + # Interactive: pick region in a browser UI + generator.annotate() + + # Then compute the best grasp pose + pose, open_length = generator.get_grasp_poses(object_pose, approach_dir) + """ + + def __init__( + self, + vertices: torch.Tensor, + triangles: torch.Tensor, + cfg: GraspGeneratorCfg = GraspGeneratorCfg(), + gripper_collision_cfg: GripperCollisionCfg = GripperCollisionCfg(), + ) -> None: + """Initialize the GraspGenerator with the given mesh vertices, triangles, and configuration. + Args: + vertices (torch.Tensor): A tensor of shape (V, 3) representing the vertex positions of the mesh. + triangles (torch.Tensor): A tensor of shape (F, 3) representing the triangle indices of the mesh. + cfg (GraspGeneratorCfg, optional): Configuration for the grasp annotator. Defaults to GraspGeneratorCfg(). + """ + self.device = vertices.device + self.vertices = vertices + self.triangles = triangles + self.mesh = trimesh.Trimesh( + vertices=vertices.to("cpu").numpy(), + faces=triangles.to("cpu").numpy(), + process=False, + force="mesh", + ) + self._collision_checker = GripperCollisionChecker( + object_mesh_verts=vertices, + object_mesh_faces=triangles, + cfg=gripper_collision_cfg, + ) + self.cfg = cfg + self._antipodal_sampler = AntipodalSampler(cfg=cfg.antipodal_sampler_cfg) + self._hit_point_pairs: torch.Tensor | None = None + + # Load cached antipodal pairs for the whole mesh if available. + cache_path = self._get_cache_dir(self.vertices, self.triangles) + if os.path.exists(cache_path): + logger.log_info(f"Found cached antipodal pairs at {cache_path}. Loading.") + self._hit_point_pairs = torch.tensor( + np.load(cache_path), dtype=torch.float32, device=self.device + ) + + def generate( + self, + vertex_indices: torch.Tensor | None = None, + face_indices: torch.Tensor | None = None, + ) -> torch.Tensor: + """ + Generate antipodal point pairs for grasping on the given mesh region. + + Exactly one of ``vertex_indices`` or ``face_indices`` must be provided + to define the grasp region. When both are ``None``, the whole mesh is + used. + + Results are cached to disk. + + Args: + vertex_indices: 1-D ``torch.Tensor`` of vertex indices defining the + grasp region. + face_indices: 1-D ``torch.Tensor`` of face indices defining the + grasp region. + + Raises: + ValueError: If both ``vertex_indices`` and ``face_indices`` are + provided at the same time. + + Returns: + torch.Tensor: A tensor of shape ``(N, 2, 3)`` representing N + antipodal point pairs. Each pair consists of a hit point and + its corresponding surface point. + """ + if vertex_indices is not None and face_indices is not None: + raise ValueError( + "Only one of vertex_indices or face_indices should be provided, not both." + ) + + if vertex_indices is None and face_indices is None: + sub_vertices = self.vertices + sub_faces = self.triangles + else: + if face_indices is not None: + face_idx_np = face_indices.cpu().numpy() + else: + vertex_idx_np = vertex_indices.cpu().numpy() + vertex_mask = np.zeros(self.mesh.vertices.shape[0], dtype=bool) + vertex_mask[vertex_idx_np] = True + face_all = cast(np.ndarray, self.mesh.faces) + face_idx_np = np.flatnonzero(np.all(vertex_mask[face_all], axis=1)) + ( + _, + _, + sub_vertices_np, + sub_faces_np, + ) = GraspGenerator._extract_selection_from_faces( + self.mesh, face_idx_np, self.cfg.use_largest_connected_component + ) + if sub_vertices_np is None: + return torch.empty(0, 2, 3, dtype=torch.float32, device=self.device) + sub_vertices = torch.as_tensor( + sub_vertices_np, dtype=torch.float32, device=self.device + ) + sub_faces = torch.as_tensor( + sub_faces_np, dtype=torch.int64, device=self.device + ) + + cache_path = self._get_cache_dir(sub_vertices, sub_faces) + if os.path.exists(cache_path): + logger.log_info(f"Found cached antipodal pairs at {cache_path}") + return torch.tensor( + np.load(cache_path), dtype=torch.float32, device=self.device + ) + + self._hit_point_pairs = self._antipodal_sampler.sample(sub_vertices, sub_faces) + self._save_cache(cache_path, self._hit_point_pairs) + return self._hit_point_pairs + + def annotate(self) -> torch.Tensor: + """Annotate antipodal grasp region on the mesh and return sampled antipodal point pairs. + + Returns: + torch.Tensor: A tensor of shape (N, 2, 3) representing N antipodal point pairs. + Each pair consists of a hit point and its corresponding surface point. + """ + + logger.log_info( + f"[Viser] *****Annotate grasp region in http://localhost:{self.cfg.viser_port}" + ) + + server = viser.ViserServer(port=self.cfg.viser_port) + server.gui.configure_theme(brand_color=(130, 0, 150)) + server.scene.set_up_direction("+z") + + mesh_handle = server.scene.add_mesh_trimesh(name="/mesh", mesh=self.mesh) + selected_overlay: viser.GlbHandle | None = None + sel_vertex_indices: np.ndarray | None = None + sel_face_indices: np.ndarray | None = None + sel_vertices: np.ndarray | None = None + sel_faces: np.ndarray | None = None + + hit_point_pairs = None + return_flag = False + + @server.on_client_connect + def _(client: viser.ClientHandle) -> None: + nonlocal mesh_handle + nonlocal selected_overlay + nonlocal sel_vertex_indices + nonlocal sel_face_indices + nonlocal sel_vertices + nonlocal sel_faces + + # client.camera.position = np.array([0.0, 0.0, -0.5]) + # client.camera.wxyz = np.array([1.0, 0.0, 0.0, 0.0]) + + select_button = client.gui.add_button( + "Rect Select Region", icon=viser.Icon.PAINT + ) + confirm_button = client.gui.add_button("Confirm Selection") + + @select_button.on_click + def _(_evt: viser.GuiEvent) -> None: + select_button.disabled = True + + @client.scene.on_pointer_event(event_type="rect-select") + def _(event: viser.ScenePointerEvent) -> None: + nonlocal mesh_handle + nonlocal selected_overlay + nonlocal sel_vertex_indices + nonlocal sel_face_indices + nonlocal sel_vertices + nonlocal sel_faces + nonlocal hit_point_pairs + client.scene.remove_pointer_callback() + + proj, depth = GraspGenerator._project_vertices_to_screen( + cast(np.ndarray, self.mesh.vertices), + mesh_handle, + event.client.camera, + ) + + lower = np.minimum( + np.array(event.screen_pos[0]), np.array(event.screen_pos[1]) + ) + upper = np.maximum( + np.array(event.screen_pos[0]), np.array(event.screen_pos[1]) + ) + vertex_mask = ((proj >= lower) & (proj <= upper)).all(axis=1) & ( + depth > 1e-6 + ) + + ( + sel_vertex_indices, + sel_face_indices, + sel_vertices, + sel_faces, + ) = GraspGenerator._extract_selection_from_vertex_mask( + self.mesh, vertex_mask, self.cfg.use_largest_connected_component + ) + if sel_vertices is None: + logger.log_warning("[Selection] No vertices selected.") + return + + color_mesh = self.mesh.copy() + vertex_colors = np.tile( + np.array([[0.85, 0.85, 0.85, 1.0]]), + (self.mesh.vertices.shape[0], 1), + ) + vertex_colors[sel_vertex_indices] = np.array( + [0.56, 0.17, 0.92, 1.0] + ) + color_mesh.visual.vertex_colors = vertex_colors # type: ignore + mesh_handle = server.scene.add_mesh_trimesh( + name="/mesh", mesh=color_mesh + ) + + if selected_overlay is not None: + selected_overlay.remove() + selected_mesh = trimesh.Trimesh( + vertices=sel_vertices, + faces=sel_faces, + process=False, + ) + selected_mesh.visual.face_colors = (0.9, 0.2, 0.2, 0.65) # type: ignore + selected_overlay = server.scene.add_mesh_trimesh( + name="/selected", mesh=selected_mesh + ) + logger.log_info( + f"[Selection] Selected {sel_vertex_indices.size} vertices and {sel_face_indices.size} faces." + ) + + hit_point_pairs = self._antipodal_sampler.sample( + torch.tensor(sel_vertices, device=self.device), + torch.tensor(sel_faces, device=self.device), + ) + + # for visualization only + extended_hit_point_pairs = GraspGenerator._extend_hit_point_pairs( + hit_point_pairs + ) + server.scene.add_line_segments( + name="/antipodal_pairs", + points=extended_hit_point_pairs.to("cpu").numpy(), + colors=(20, 200, 200), + line_width=1.5, + ) + + @client.scene.on_pointer_callback_removed + def _() -> None: + select_button.disabled = False + + @confirm_button.on_click + def _(_evt: viser.GuiEvent) -> None: + nonlocal return_flag + if sel_vertices is None: + logger.log_warning("[Selection] No vertex selected.") + return + else: + logger.log_info( + f"[Selection] {sel_vertices.shape[0]}vertices selected. Generating antipodal point pairs." + ) + return_flag = True + + while True: + if return_flag: + if hit_point_pairs is not None: + self._hit_point_pairs = hit_point_pairs + cache_path = self._get_cache_dir(self.vertices, self.triangles) + self._save_cache(cache_path, hit_point_pairs) + break + time.sleep(0.5) + return self._hit_point_pairs + + def _get_cache_dir(self, vertices: torch.Tensor, triangles: torch.Tensor): + vert_bytes = vertices.to("cpu").numpy().tobytes() + face_bytes = triangles.to("cpu").numpy().tobytes() + md5_hash = hashlib.md5(vert_bytes + face_bytes).hexdigest() + cache_path = os.path.join( + GRASP_ANNOTATOR_CACHE_DIR, f"antipodal_cache_{md5_hash}.npy" + ) + return cache_path + + def _save_cache(self, cache_path: str, hit_point_pairs: torch.Tensor): + np.save(cache_path, hit_point_pairs.cpu().numpy().astype(np.float32)) + + @staticmethod + def _extend_hit_point_pairs(hit_point_pairs: torch.Tensor): + origin_points = hit_point_pairs[:, 0, :] + hit_points = hit_point_pairs[:, 1, :] + mid_points = (origin_points + hit_points) / 2 + point_diff = hit_points - origin_points + extended_origin = mid_points - 0.8 * point_diff + extended_hit = mid_points + 0.8 * point_diff + extended_point_pairs = torch.cat( + [extended_origin[:, None, :], extended_hit[:, None, :]], dim=1 + ) + return extended_point_pairs + + @staticmethod + def _project_vertices_to_screen( + vertices_mesh: np.ndarray, + mesh_handle: viser.GlbHandle, + camera: Any, + ) -> tuple[np.ndarray, np.ndarray]: + T_world_mesh = tf.SE3.from_rotation_and_translation( + tf.SO3(np.asarray(mesh_handle.wxyz)), + np.asarray(mesh_handle.position), + ) + vertices_world_h = ( + T_world_mesh.as_matrix() + @ np.hstack([vertices_mesh, np.ones((vertices_mesh.shape[0], 1))]).T + ).T + vertices_world = vertices_world_h[:, :3] + + T_camera_world = tf.SE3.from_rotation_and_translation( + tf.SO3(np.asarray(camera.wxyz)), + np.asarray(camera.position), + ).inverse() + vertices_camera_h = ( + T_camera_world.as_matrix() + @ np.hstack([vertices_world, np.ones((vertices_world.shape[0], 1))]).T + ).T + vertices_camera = vertices_camera_h[:, :3] + + fov = float(camera.fov) + aspect = float(camera.aspect) + projected = vertices_camera[:, :2] / np.maximum(vertices_camera[:, 2:3], 1e-8) + projected /= np.tan(fov / 2.0) + projected[:, 0] /= aspect + projected = (1.0 + projected) / 2.0 + return projected, vertices_camera[:, 2] + + @staticmethod + def _extract_selection_from_vertex_mask( + mesh: trimesh.Trimesh, + vertex_mask: np.ndarray, + largest_component: bool, + ) -> tuple[ + np.ndarray | None, np.ndarray | None, np.ndarray | None, np.ndarray | None + ]: + """Extract a sub-mesh from *mesh* using a per-vertex boolean mask. + + Args: + mesh: The source mesh. + vertex_mask: Boolean array of shape ``(V,)`` indicating which + vertices are selected. + largest_component: If ``True``, keep only the largest connected + component among the selected faces. + + Returns: + A tuple ``(vertex_indices, face_indices, sub_vertices, sub_faces)`` + where ``sub_vertices`` and ``sub_faces`` define the extracted + sub-mesh with remapped indices. Returns ``(None, None, None, None)`` + if no faces are selected. + """ + faces = cast(np.ndarray, mesh.faces) + face_mask = np.all(vertex_mask[faces], axis=1) + face_indices = np.flatnonzero(face_mask) + if face_indices.size == 0: + return None, None, None, None + if largest_component: + face_indices = GraspGenerator._largest_connected_face_component( + mesh, face_indices + ) + if face_indices.size == 0: + return None, None, None, None + return GraspGenerator._build_sub_mesh(mesh, face_indices) + + @staticmethod + def _extract_selection_from_faces( + mesh: trimesh.Trimesh, + face_indices: np.ndarray, + largest_component: bool, + ) -> tuple[ + np.ndarray | None, np.ndarray | None, np.ndarray | None, np.ndarray | None + ]: + """Extract a sub-mesh from *mesh* using face indices. + + Args: + mesh: The source mesh. + face_indices: Array of face indices to include. + largest_component: If ``True``, keep only the largest connected + component among the selected faces. + + Returns: + Same as :meth:`_extract_selection_from_vertex_mask`. + """ + if face_indices.size == 0: + return None, None, None, None + if largest_component: + face_indices = GraspGenerator._largest_connected_face_component( + mesh, face_indices + ) + if face_indices.size == 0: + return None, None, None, None + return GraspGenerator._build_sub_mesh(mesh, face_indices) + + @staticmethod + def _build_sub_mesh( + mesh: trimesh.Trimesh, + face_indices: np.ndarray, + ) -> tuple[np.ndarray, np.ndarray, np.ndarray, np.ndarray]: + """Build a sub-mesh with remapped vertex indices from selected faces. + + Returns: + ``(vertex_indices, face_indices, sub_vertices, sub_faces)`` + """ + faces = cast(np.ndarray, mesh.faces) + selected_face_vertices = faces[face_indices] + vertex_indices = np.unique(selected_face_vertices.reshape(-1)) + + old_to_new = np.full(mesh.vertices.shape[0], -1, dtype=np.int32) + old_to_new[vertex_indices] = np.arange(vertex_indices.size, dtype=np.int32) + + sub_vertices = np.asarray(mesh.vertices)[vertex_indices] + sub_faces = np.asarray(old_to_new)[selected_face_vertices] + + return vertex_indices, face_indices, sub_vertices, sub_faces + + @staticmethod + def _largest_connected_face_component( + mesh: trimesh.Trimesh, + face_ids: np.ndarray, + ) -> np.ndarray: + """Return the face indices of the largest connected component.""" + if face_ids.size <= 1: + return face_ids + + face_id_set = set(face_ids.tolist()) + parent: dict[int, int] = {int(face_id): int(face_id) for face_id in face_ids} + + def find(x: int) -> int: + root = x + while parent[root] != root: + root = parent[root] + while parent[x] != x: + x_parent = parent[x] + parent[x] = root + x = x_parent + return root + + def union(a: int, b: int) -> None: + ra, rb = find(a), find(b) + if ra != rb: + parent[rb] = ra + + face_adjacency = cast(np.ndarray, mesh.face_adjacency) + for face_a, face_b in face_adjacency: + if int(face_a) in face_id_set and int(face_b) in face_id_set: + union(int(face_a), int(face_b)) + + groups: dict[int, list[int]] = {} + for face_id in face_ids: + root = find(int(face_id)) + groups.setdefault(root, []).append(int(face_id)) + + largest_group = max(groups.values(), key=len) + return np.array(largest_group, dtype=np.int32) + + @staticmethod + def _apply_transform(points: torch.Tensor, transform: torch.Tensor) -> torch.Tensor: + r = transform[:3, :3] + t = transform[:3, 3] + return points @ r.T + t + + def get_grasp_poses( + self, + object_pose: torch.Tensor, + approach_direction: torch.Tensor, + visualize_collision: bool = False, + visualize_pose: bool = False, + ) -> tuple[torch.Tensor, torch.Tensor]: + """Get grasp pose given approach direction. + + Uses the antipodal point pairs stored in ``self._hit_point_pairs`` + (populated by :meth:`generate` or :meth:`annotate`). + + TODO: + 1. Support Top-k grasp poses selection. + 2. Support more selection criteria. + + Args: + object_pose: ``(4, 4)`` homogeneous transformation matrix + representing the pose of the object in the world frame. + approach_direction: ``(3,)`` unit vector representing the desired + approach direction of the gripper in the world frame. + visualize_collision: If ``True``, enable visual collision checking. + visualize_pose: If ``True``, visualize the best grasp pose using Open3D + after computation. + + Returns: + A tuple ``(best_grasp_pose, best_open_length)`` where + ``best_grasp_pose`` is a ``(4, 4)`` homogeneous matrix and + ``best_open_length`` is a scalar. + + Raises: + RuntimeError: If :meth:`generate` or :meth:`annotate` has not + been called yet. + """ + if self._hit_point_pairs is None: + raise RuntimeError( + "No antipodal point pairs available. " + "Call generate() or annotate() first." + ) + origin_points = self._hit_point_pairs[:, 0, :] + hit_points = self._hit_point_pairs[:, 1, :] + origin_points_ = self._apply_transform(origin_points, object_pose) + hit_points_ = self._apply_transform(hit_points, object_pose) + centers = (origin_points_ + hit_points_) / 2 + + mesh_vert_transformed = self._apply_transform(self.vertices, object_pose) + mesh_center = mesh_vert_transformed.mean(dim=0) + + # filter perpendicular antipodal point + grasp_x = F.normalize(hit_points_ - origin_points_, dim=-1) + cos_angle = torch.clamp((grasp_x * approach_direction).sum(dim=-1), -1.0, 1.0) + positive_angle = torch.abs(torch.acos(cos_angle)) + valid_mask = ( + positive_angle - torch.pi / 2 + ).abs() <= self.cfg.max_deviation_angle + valid_grasp_x = grasp_x[valid_mask] + valid_centers = centers[valid_mask] + + # compute grasp poses using antipodal point pairs and approach direction + valid_grasp_poses = GraspGenerator._grasp_pose_from_approach_direction( + valid_grasp_x, approach_direction, valid_centers + ) + valid_open_lengths = torch.norm( + origin_points_[valid_mask] - hit_points_[valid_mask], dim=-1 + ) + # select non-collide grasp poses + is_colliding, max_penetration = self._collision_checker.query( + object_pose, + valid_grasp_poses, + valid_open_lengths, + is_visual=visualize_collision, + collision_threshold=0.0, + ) + # get best grasp pose + valid_grasp_poses = valid_grasp_poses[~is_colliding] + valid_open_lengths = valid_open_lengths[~is_colliding] + valid_centers = valid_centers[~is_colliding] + valid_grasp_x = F.normalize(valid_grasp_poses[:, :3, 0], dim=-1) + + cos_angle = torch.clamp( + (valid_grasp_x * approach_direction).sum(dim=-1), -1.0, 1.0 + ) + positive_angle = torch.abs(torch.acos(cos_angle)) + angle_cost = torch.abs(positive_angle - 0.5 * torch.pi) / (0.5 * torch.pi) + center_distance = torch.norm(valid_centers - mesh_center, dim=-1) + center_cost = center_distance / center_distance.max() + length_cost = 1 - valid_open_lengths / valid_open_lengths.max() + total_cost = 0.3 * angle_cost + 0.3 * length_cost + 0.4 * center_cost + best_idx = torch.argmin(total_cost) + best_grasp_pose = valid_grasp_poses[best_idx] + best_open_length = valid_open_lengths[best_idx] + if visualize_pose: + self.visualize_grasp_pose( + obj_pose=object_pose, + grasp_pose=best_grasp_pose, + open_length=best_open_length.item(), + ) + return best_grasp_pose, best_open_length + + @staticmethod + def _grasp_pose_from_approach_direction( + grasp_x: torch.Tensor, approach_direction: torch.Tensor, center: torch.Tensor + ): + approach_direction_repeat = approach_direction[None, :].repeat( + grasp_x.shape[0], 1 + ) + grasp_y = torch.cross(approach_direction_repeat, grasp_x, dim=-1) + grasp_y = F.normalize(grasp_y, dim=-1) + grasp_z = torch.cross(grasp_x, grasp_y, dim=-1) + grasp_z = F.normalize(grasp_z, dim=-1) + grasp_poses = ( + torch.eye(4, device=grasp_x.device, dtype=torch.float32) + .unsqueeze(0) + .repeat(grasp_x.shape[0], 1, 1) + ) + grasp_poses[:, :3, 0] = grasp_x + grasp_poses[:, :3, 1] = grasp_y + grasp_poses[:, :3, 2] = grasp_z + grasp_poses[:, :3, 3] = center + return grasp_poses + + def visualize_grasp_pose( + self, + obj_pose: torch.Tensor, + grasp_pose: torch.Tensor, + open_length: float, + ): + mesh = o3d.geometry.TriangleMesh( + vertices=o3d.utility.Vector3dVector(self.vertices.to("cpu").numpy()), + triangles=o3d.utility.Vector3iVector(self.triangles.to("cpu").numpy()), + ) + mesh.compute_vertex_normals() + mesh.paint_uniform_color([0.3, 0.6, 0.3]) + mesh.transform(obj_pose.to("cpu").numpy()) + vertices_ = torch.tensor( + np.asarray(mesh.vertices), + device=self.vertices.device, + dtype=self.vertices.dtype, + ) + mesh_scale = (vertices_.max(dim=0)[0] - vertices_.min(dim=0)[0]).max().item() + groud_plane = o3d.geometry.TriangleMesh.create_cylinder( + radius=mesh_scale, height=0.01 * mesh_scale + ) + groud_plane.compute_vertex_normals() + center = vertices_.mean(dim=0) + z_sim = vertices_.min(dim=0)[0][2].item() + groud_plane.translate( + (center[0].item(), center[1].item(), z_sim - 0.005 * mesh_scale) + ) + + draw_thickness = 0.02 * mesh_scale + draw_length = 0.3 * mesh_scale + grasp_finger1 = o3d.geometry.TriangleMesh.create_box( + draw_thickness, draw_thickness, draw_length + ) + grasp_finger1.translate( + (-0.5 * draw_thickness, -0.5 * draw_thickness, -0.5 * draw_length) + ) + grasp_finger2 = o3d.geometry.TriangleMesh.create_box( + draw_thickness, draw_thickness, draw_length + ) + grasp_finger2.translate( + (-0.5 * draw_thickness, -0.5 * draw_thickness, -0.5 * draw_length) + ) + grasp_finger1.translate((-open_length / 2, 0, -0.25 * draw_length)) + grasp_finger2.translate((open_length / 2, 0, -0.25 * draw_length)) + grasp_root1 = o3d.geometry.TriangleMesh.create_box( + open_length, draw_thickness, draw_thickness + ) + grasp_root1.translate( + (-open_length / 2, -0.5 * draw_thickness, -0.5 * draw_thickness) + ) + grasp_root1.translate((0, 0, -0.75 * draw_length)) + grasp_root2 = o3d.geometry.TriangleMesh.create_box( + draw_thickness, draw_thickness, draw_length + ) + grasp_root2.translate( + (-0.5 * draw_thickness, -0.5 * draw_thickness, -0.5 * draw_length) + ) + grasp_root2.translate((0, 0, -1.25 * draw_length)) + + grasp_visual = grasp_finger1 + grasp_finger2 + grasp_root1 + grasp_root2 + grasp_visual.paint_uniform_color([0.8, 0.2, 0.8]) + grasp_visual.transform(grasp_pose.to("cpu").numpy()) + o3d.visualization.draw_geometries( + [grasp_visual, mesh, groud_plane], + window_name="Grasp Pose Visualization", + mesh_show_back_face=True, + ) diff --git a/embodichain/toolkits/graspkit/pg_grasp/antipodal_sampler.py b/embodichain/toolkits/graspkit/pg_grasp/antipodal_sampler.py new file mode 100644 index 00000000..5e4b7594 --- /dev/null +++ b/embodichain/toolkits/graspkit/pg_grasp/antipodal_sampler.py @@ -0,0 +1,253 @@ +# ---------------------------------------------------------------------------- +# Copyright (c) 2021-2026 DexForce Technology Co., Ltd. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# ---------------------------------------------------------------------------- + +import torch +import torch.nn.functional as F +import numpy as np +import open3d as o3d +import open3d.core as o3c +from embodichain.utils import configclass +from embodichain.utils import logger + +__all__ = ["AntipodalSamplerCfg", "AntipodalSampler"] + + +@configclass +class AntipodalSamplerCfg: + """Configuration for AntipodalSampler.""" + + n_sample: int = 20000 + """surface point sample number""" + + max_angle: float = np.pi / 12 + """maximum angle (in radians) to randomly disturb the ray direction for antipodal point sampling, + used to increase the diversity of sampled antipodal points. Note that setting max_angle to 0 will + disable the random disturbance and sample antipodal points strictly along the surface normals, + which may result in less diverse antipodal points and may not be ideal for all objects or grasping + scenarios. + """ + + max_length: float = 0.1 + """maximum gripper open width, used to filter out antipodal points that are too far apart to be grasped""" + + min_length: float = 0.001 + """minimum gripper open width, used to filter out antipodal points that are too close to be grasped""" + + +class AntipodalSampler: + """AntipodalSampler samples antipodal point pairs on a given mesh. It uses Open3D's raycasting functionality to find points on the mesh that are visible along the negative normal direction from uniformly sampled points on the mesh surface. The sampler can also apply a random disturbance to the ray direction to increase the diversity of sampled antipodal points. The resulting antipodal point pairs can be used for grasp generation and annotation tasks.""" + + def __init__( + self, + cfg: AntipodalSamplerCfg = AntipodalSamplerCfg(), + ): + self.mesh: o3d.t.geometry.TriangleMesh | None = None + self.cfg = cfg + + def sample(self, vertices: torch.Tensor, faces: torch.Tensor) -> torch.Tensor: + """Get sample Antipodal point pair + + Args: + vertices: [V, 3] vertex positions of the mesh + faces: [F, 3] triangle indices of the mesh + + Returns: + hit_point_pairs: [N, 2, 3] tensor of N antipodal point pairs. Each pair consists of a hit point and its corresponding surface point. + """ + # update mesh + self.mesh = o3d.t.geometry.TriangleMesh() + self.mesh.vertex.positions = o3c.Tensor( + vertices.to("cpu").numpy(), dtype=o3c.float32 + ) + self.mesh.triangle.indices = o3c.Tensor( + faces.to("cpu").numpy(), dtype=o3c.int32 + ) + self.mesh.compute_vertex_normals() + # sample points and normals + sample_pcd = self.mesh.sample_points_uniformly( + number_of_points=self.cfg.n_sample + ) + sample_points = torch.tensor( + sample_pcd.point.positions.numpy(), + device=vertices.device, + dtype=vertices.dtype, + ) + sample_normals = torch.tensor( + sample_pcd.point.normals.numpy(), + device=vertices.device, + dtype=vertices.dtype, + ) + # generate rays + ray_direc = -sample_normals + ray_origin = ( + sample_points + 1e-3 * ray_direc + ) # Offset ray origin slightly along the normal to avoid self-intersection + disturb_direc = AntipodalSampler._random_rotate_unit_vectors( + ray_direc, max_angle=self.cfg.max_angle + ) + ray_origin = torch.vstack([ray_origin, ray_origin]) + ray_direc = torch.vstack([ray_direc, disturb_direc]) + # casting + return self._get_raycast_result( + ray_origin, + ray_direc, + surface_origin=torch.vstack([sample_points, sample_points]), + ) + + def _get_raycast_result( + self, + ray_origin: torch.Tensor, + ray_direc: torch.Tensor, + surface_origin: torch.Tensor, + ): + if ray_origin.ndim != 2 or ray_origin.shape[-1] != 3: + raise ValueError("ray_origin must have shape [N, 3]") + if ray_direc.ndim != 2 or ray_direc.shape[-1] != 3: + raise ValueError("ray_direc must have shape [N, 3]") + if ray_origin.shape[0] != ray_direc.shape[0]: + raise ValueError( + "ray_origin and ray_direc must have the same number of rays" + ) + if ray_origin.shape[0] != surface_origin.shape[0]: + raise ValueError( + "ray_origin and surface_origin must have the same number of rays" + ) + + scene = o3d.t.geometry.RaycastingScene() + scene.add_triangles(self.mesh) + + rays = torch.cat([ray_origin, ray_direc], dim=-1) + rays_o3d = o3c.Tensor(rays.detach().to("cpu").numpy(), dtype=o3c.float32) + + ans = scene.cast_rays(rays_o3d) + t_hit = torch.from_numpy(ans["t_hit"].numpy()).to( + device=ray_origin.device, dtype=ray_origin.dtype + ) + hit_mask = torch.logical_and( + t_hit > self.cfg.min_length, t_hit < self.cfg.max_length + ) + hit_points = ray_origin[hit_mask] + t_hit[hit_mask, None] * ray_direc[hit_mask] + hit_origins = surface_origin[hit_mask] + hit_point_pairs = torch.cat( + [hit_points[:, None, :], hit_origins[:, None, :]], dim=1 + ) + hit_point_pairs = hit_point_pairs.to(dtype=torch.float32) + return hit_point_pairs + + @staticmethod + def _random_rotate_unit_vectors( + vectors: torch.Tensor, + max_angle: float, + degrees: bool = False, + eps: float = 1e-8, + ) -> torch.Tensor: + """ + Apply random small rotations to a batch of unit vectors [N, 3]. + + Args: + vectors: [N, 3], unit vectors + max_angle: Maximum rotation angle + degrees: If True, `max_angle` is given in degrees + eps: Numerical stability constant + + Returns: + rotated: [N, 3], rotated unit vectors + """ + assert vectors.ndim == 2 and vectors.shape[-1] == 3, "vectors must be [N, 3]" + + v = F.normalize(vectors, dim=-1) + + if degrees: + max_angle = torch.deg2rad( + torch.tensor(max_angle, dtype=v.dtype, device=v.device) + ).item() + + n = v.shape[0] + + # 1) Generate a random direction for each vector + # then project it onto the plane perpendicular to v to get the rotation axis k + rand_dir = torch.randn_like(v) + eps + proj = (rand_dir * v).sum(dim=-1, keepdim=True) * v + k = rand_dir - proj + k = F.normalize(k, dim=-1) + + # 2) Sample rotation angles in the range [eps, max_angle] + theta = ( + torch.rand(n, 1, device=v.device, dtype=v.dtype) * (max_angle - eps) + eps + ) + + # 3) Rodrigues' rotation formula + # R(v) = v*cosθ + (k×v)*sinθ + k*(k·v)*(1-cosθ) + # Since k ⟂ v, the last term is theoretically 0, but keeping the general formula is more robust + cos_t = torch.cos(theta) + sin_t = torch.sin(theta) + + kv = (k * v).sum(dim=-1, keepdim=True) + rotated = v * cos_t + torch.cross(k, v, dim=-1) * sin_t + k * kv * (1.0 - cos_t) + + return F.normalize(rotated, dim=-1) + + def visualize(self, hit_point_pairs: torch.Tensor): + if self.mesh is None: + logger.log_warning("Mesh is not initialized. Cannot visualize.") + return + + if hit_point_pairs.shape[0] == 0: + raise ValueError("No point pairs to visualize") + origin_points = hit_point_pairs[:, 0, :] + hit_points = hit_point_pairs[:, 1, :] + + origin_points_np = origin_points.to("cpu").numpy() + hit_points_np = hit_points.detach().to("cpu").numpy() + + n_pairs = hit_point_pairs.shape[0] + line_indices = np.stack( + [np.arange(n_pairs), np.arange(n_pairs) + n_pairs], axis=1 + ) + + mesh_legacy = self.mesh.to_legacy() + mesh_legacy.compute_vertex_normals() + mesh_legacy.paint_uniform_color([0.8, 0.8, 0.8]) + + origin_pcd = o3d.geometry.PointCloud() + origin_pcd.points = o3d.utility.Vector3dVector(origin_points_np) + origin_pcd.colors = o3d.utility.Vector3dVector( + np.tile(np.array([[0.1, 0.4, 1.0]]), (n_pairs, 1)) + ) + + hit_pcd = o3d.geometry.PointCloud() + hit_pcd.points = o3d.utility.Vector3dVector(hit_points_np) + hit_pcd.colors = o3d.utility.Vector3dVector( + np.tile(np.array([[1.0, 0.2, 0.2]]), (n_pairs, 1)) + ) + + line_set = o3d.geometry.LineSet() + mid_points = (origin_points_np + hit_points_np) / 2 + point_diff = hit_points_np - origin_points_np + draw_origin = mid_points - 0.6 * point_diff + draw_end = mid_points + 0.6 * point_diff + draw_pointpair = np.concatenate([draw_origin, draw_end], axis=0) + line_set.points = o3d.utility.Vector3dVector(draw_pointpair) + line_set.lines = o3d.utility.Vector2iVector(line_indices) + line_set.colors = o3d.utility.Vector3dVector( + np.tile(np.array([[0.2, 0.9, 0.2]]), (n_pairs, 1)) + ) + + o3d.visualization.draw_geometries( + [mesh_legacy, origin_pcd, hit_pcd, line_set], + window_name="Antipodal Point Pairs", + mesh_show_back_face=True, + ) diff --git a/embodichain/toolkits/graspkit/pg_grasp/collision_checker.py b/embodichain/toolkits/graspkit/pg_grasp/collision_checker.py new file mode 100644 index 00000000..fcbfb850 --- /dev/null +++ b/embodichain/toolkits/graspkit/pg_grasp/collision_checker.py @@ -0,0 +1,419 @@ +# ---------------------------------------------------------------------------- +# Copyright (c) 2021-2026 DexForce Technology Co., Ltd. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# ---------------------------------------------------------------------------- + +import trimesh +import numpy as np +import torch +import warp as wp +import time +import hashlib +import os +import pickle +import open3d as o3d + +from typing import List, Tuple, Union +from dexsim.kit.meshproc import convex_decomposition_coacd + +from embodichain.utils.warp import convex_signed_distance_kernel +from embodichain.utils.device_utils import standardize_device_string +from embodichain.utils.math import transform_points_mat +from embodichain.utils import configclass + +__all__ = ["ConvexCollisionCheckerCfg", "ConvexCollisionChecker"] + + +@configclass +class ConvexCollisionCheckerCfg: + """Configuration for ConvexCollisionChecker.""" + + collision_threshold: float = 0.0 + """ Collision threshold in meters. A point is considered colliding if its signed distance to the hull interior is <= this threshold. This allows for a margin of error in collision checking, where a small positive threshold can be used to consider points near the surface as colliding, and a small negative threshold can be used to allow for slight penetration without considering it a collision.""" + + n_query_mesh_samples: int = 4096 + """ Number of points to sample from the query mesh surface for collision checking. A higher number of samples can provide a more accurate collision check at the cost of increased computation time. The optimal number may depend on the complexity of the mesh and the required precision of collision detection.""" + + debug: bool = False + """ Whether to visualize the collision checking results for debugging purposes. If set to True, the code will generate visualizations of the query points colored by their collision status (e.g., red for colliding points and green for non-colliding points) along with the original mesh. This can help in understanding and verifying the collision checking process, especially during development and testing.""" + + +class ConvexCollisionChecker: + """ConvexCollisionChecker performs efficient collision checking between a batch of query point clouds and a convex decomposition of a mesh. The convex decomposition is represented by plane equations of the convex hulls, which are precomputed and cached for efficiency. The collision checking is done by computing the signed distance from each query point to the convex hulls using the plane equations, and determining if any points are colliding based on a specified collision threshold. This class can be used""" + + def __init__( + self, + base_mesh_verts: torch.Tensor, + base_mesh_faces: torch.Tensor, + max_decomposition_hulls: int = 32, + ): + """Initialize the ConvexCollisionChecker by performing convex decomposition on the input mesh and extracting plane equations for the convex hulls. The plane equations are cached to disk to avoid redundant computation in future runs. + + Args: + base_mesh_verts: [N, 3] vertex positions of the input mesh. + base_mesh_faces: [M, 3] triangle indices of the input mesh. + max_decomposition_hulls: maximum number of convex hulls to decompose into. A higher number allows for a more accurate approximation of the original mesh but increases computation time and memory usage. The optimal number may depend on the complexity of the mesh and the required precision of collision checking. + """ + from embodichain.lab.sim import CONVEX_DECOMP_DIR + + if not os.path.isdir(CONVEX_DECOMP_DIR): + os.makedirs(CONVEX_DECOMP_DIR, exist_ok=True) + self.device = base_mesh_verts.device + base_mesh_verts_np = base_mesh_verts.cpu().numpy() + base_mesh_faces_np = base_mesh_faces.cpu().numpy() + mesh_hash = hashlib.md5( + (base_mesh_verts_np.tobytes() + base_mesh_faces_np.tobytes()) + ).hexdigest() + + # for visualization + self.mesh = o3d.geometry.TriangleMesh( + vertices=o3d.utility.Vector3dVector(base_mesh_verts_np), + triangles=o3d.utility.Vector3iVector(base_mesh_faces_np), + ) + self.mesh.compute_vertex_normals() + + self.cache_path = os.path.join( + CONVEX_DECOMP_DIR, f"{mesh_hash}_{max_decomposition_hulls}.pkl" + ) + + if not os.path.isfile(self.cache_path): + # [n_convex, n_max_faces, 4]: plane equations, normals(3) and offsets(1), padded with zeros if a hull has less than n_max_faces + # [n_convex, ]: number of faces for each convex hull + + # generate convex hulls and extract plane equations, then cache to disk + plane_equations_np = ConvexCollisionChecker._compute_plane_equations( + base_mesh_verts_np, base_mesh_faces_np, max_decomposition_hulls + ) + # pack as a single tensor + n_convex = len(plane_equations_np) + n_max_equation = max(len(normals) for normals, _ in plane_equations_np) + plane_equations = torch.zeros( + size=(n_convex, n_max_equation, 4), + dtype=torch.float32, + device=self.device, + ) + plane_equations_counts = torch.zeros( + n_convex, dtype=torch.int32, device=self.device + ) + for i in range(n_convex): + n_equation = plane_equations_np[i][0].shape[0] + # plane normals + plane_equations[i, :n_equation, :3] = torch.tensor( + plane_equations_np[i][0], device=self.device + ) + # plane offsets + plane_equations[i, :n_equation, 3] = torch.tensor( + plane_equations_np[i][1], device=self.device + ) + plane_equations_counts[i] = n_equation + self.plane_equations = { + "plane_equations": plane_equations, + "plane_equation_counts": plane_equations_counts, + } + pickle.dump(self.plane_equations, open(self.cache_path, "wb")) + else: + self.plane_equations = pickle.load(open(self.cache_path, "rb")) + self.plane_equations["plane_equations"] = self.plane_equations[ + "plane_equations" + ].to(self.device) + self.plane_equations["plane_equation_counts"] = self.plane_equations[ + "plane_equation_counts" + ].to(self.device) + + @staticmethod + def batch_point_convex_query( + plane_equations: torch.Tensor, + plane_equation_counts: torch.Tensor, + batch_points: torch.Tensor, + device: torch.device, + collision_threshold: float = -0.003, + ): + # always use cuda for batch grasp pose query + is_cpu = device == torch.device("cpu") + if is_cpu: + plane_equations_wp = wp.from_torch(plane_equations.to("cuda")) + plane_equation_counts_wp = wp.from_torch(plane_equation_counts.to("cuda")) + batch_points_wp = wp.from_torch(batch_points.to("cuda")) + else: + plane_equations_wp = wp.from_torch(plane_equations) + plane_equation_counts_wp = wp.from_torch(plane_equation_counts) + batch_points_wp = wp.from_torch(batch_points) + + if is_cpu: + wp_device = standardize_device_string(torch.device("cuda")) + else: + wp_device = standardize_device_string(device) + n_pose = batch_points.shape[0] + n_point = batch_points.shape[1] + n_convex = plane_equations.shape[0] + point_convex_signed_distance_wp = wp.full( + shape=(n_pose, n_point, n_convex), + value=-float("inf"), + dtype=float, + device=wp_device, + ) # [n_pose, n_point, n_convex] + wp.launch( + kernel=convex_signed_distance_kernel, + dim=(n_pose, n_point, n_convex), + inputs=(batch_points_wp, plane_equations_wp, plane_equation_counts_wp), + outputs=(point_convex_signed_distance_wp,), + device=wp_device, + ) + point_convex_signed_distance = wp.to_torch(point_convex_signed_distance_wp) + point_signed_distance = point_convex_signed_distance.min( + dim=-1 + ).values # [n_pose, n_point] + is_point_collide = point_signed_distance <= collision_threshold + if is_cpu: + return point_signed_distance.to("cpu"), is_point_collide.to("cpu") + else: + return point_signed_distance, is_point_collide + + def query_batch_points( + self, + batch_points: torch.Tensor, + collision_threshold: float = 0.0, + is_visual: bool = False, + ) -> torch.Tensor: + """Query collision status for a batch of point clouds. The collision status is determined by checking if the signed distance from any point in the cloud to the convex hulls is less than or equal to the specified collision threshold. + Args: + batch_points: [B, n_point, 3] batch of point clouds to query for collision status. + collision_threshold: Collision threshold in meters. A point is considered colliding if its signed distance to the hull interior is <= this threshold. This allows for a margin of error in collision checking, where a small positive threshold can be used to consider points near the surface as colliding, and a small negative threshold can be used to allow for slight penetration without considering it a collision. + is_visual: Whether to visualize the collision checking results for debugging purposes. If set to True, the code will generate visualizations of the query points colored by their collision status (e.g., red for colliding points and green for non-colliding points) along with the original mesh. This can help in understanding and verifying the collision checking process, especially during development and testing. + Returns: + is_pose_collide: [B, ] boolean tensor indicating whether each point cloud in the + """ + n_batch = batch_points.shape[0] + point_signed_distance, is_point_collide = ( + ConvexCollisionChecker.batch_point_convex_query( + self.plane_equations["plane_equations"], + self.plane_equations["plane_equation_counts"], + batch_points, + device=self.device, + collision_threshold=collision_threshold, + ) + ) + is_pose_collide = is_point_collide.any(dim=-1) # [B] + pose_surface_distance = point_signed_distance.min(dim=-1).values # [B] + if is_visual: + # visualize result + frame = o3d.geometry.TriangleMesh.create_coordinate_frame(size=0.1) + for i in range(n_batch): + query_points_o3d = o3d.geometry.PointCloud() + query_points_np = batch_points[i].cpu().numpy() + query_points_o3d.points = o3d.utility.Vector3dVector(query_points_np) + query_points_color = np.zeros_like(query_points_np) + query_points_color[is_point_collide[i].cpu().numpy()] = [ + 1.0, + 0, + 0, + ] # red for colliding points + query_points_color[~is_point_collide[i].cpu().numpy()] = [ + 0, + 1.0, + 0, + ] # green for non-colliding points + query_points_o3d.colors = o3d.utility.Vector3dVector(query_points_color) + o3d.visualization.draw_geometries( + [self.mesh, query_points_o3d, frame], mesh_show_back_face=True + ) + return is_pose_collide, pose_surface_distance + + def query( + self, + query_mesh_verts: torch.Tensor, + query_mesh_faces: torch.Tensor, + poses: torch.Tensor, + cfg: ConvexCollisionCheckerCfg = ConvexCollisionCheckerCfg(), + ) -> Tuple[torch.Tensor, torch.Tensor]: + query_mesh = trimesh.Trimesh( + vertices=query_mesh_verts.to("cpu").numpy(), + faces=query_mesh_faces.to("cpu").numpy(), + ) + n_query = cfg.n_query_mesh_samples + n_batch = poses.shape[0] + query_points_np = query_mesh.sample(n_query).astype(np.float32) + query_points = torch.tensor( + query_points_np, device=poses.device + ) # [n_query, 3] + penetration_result = torch.zeros(size=(n_batch, n_query), device=poses.device) + penetration_result.fill_(-float("inf")) + collision_result = torch.zeros( + size=(n_batch, n_query), dtype=torch.bool, device=poses.device + ) + collision_result.fill_(False) + for normals, offsets in self.plane_equations: + normals_torch = torch.tensor(normals, device=poses.device) + offsets_torch = torch.tensor(offsets, device=poses.device) + penetration, collides = check_collision_single_hull( + normals_torch, + offsets_torch, + transform_points_mat(query_points, poses), + cfg.collision_threshold, + ) + penetration_result = torch.max(penetration_result, penetration) + collision_result = torch.logical_or(collision_result, collides) + is_colliding = collision_result.any(dim=-1) # [B] + max_penetration = penetration_result.max(dim=-1)[0] # [B] + + if cfg.debug: + # visualize result + for i in range(n_batch): + query_points_o3d = o3d.geometry.PointCloud() + query_points_o3d.points = o3d.utility.Vector3dVector(query_points_np) + query_points_o3d.transform(poses[i].to("cpu").numpy()) + query_points_color = np.zeros_like(query_points_np) + query_points_color[collision_result[i].cpu().numpy()] = [ + 1.0, + 0, + 0, + ] # red for colliding points + query_points_color[~collision_result[i].cpu().numpy()] = [ + 0, + 1.0, + 0, + ] # green for non-colliding points + query_points_o3d.colors = o3d.utility.Vector3dVector(query_points_color) + o3d.visualization.draw_geometries( + [self.mesh, query_points_o3d], mesh_show_back_face=True + ) + return is_colliding, max_penetration + + @staticmethod + def _compute_plane_equations( + vertices: np.ndarray, faces: np.ndarray, max_decomposition_hulls: int + ) -> Tuple[np.ndarray, np.ndarray]: + """ + Convex decomposition and extract plane equations given mesh vertices and triangles. + Each convex hull is represented by its outward-facing face normals and offsets. + No padding is applied; each hull can have a different number of faces. + + Args: + vertices: [N, 3] vertex positions of the input mesh. + faces: [M, 3] triangle indices of the input mesh. + max_decomposition_hulls: maximum number of convex hulls to decompose into. + + Returns: + List of (normals_i [Ki, 3], offsets_i [Ki]) tuples, one per convex hull. + Ki is the number of faces of the i-th hull and can differ across hulls. + """ + mesh = o3d.t.geometry.TriangleMesh() + mesh.vertex.positions = o3d.core.Tensor(vertices, dtype=o3d.core.Dtype.Float32) + mesh.triangle.indices = o3d.core.Tensor(faces, dtype=o3d.core.Dtype.Int32) + is_success, out_mesh_list = convex_decomposition_coacd( + mesh, max_convex_hull_num=max_decomposition_hulls + ) + convex_vert_face_list = [] + for out_mesh in out_mesh_list: + verts = out_mesh.vertex.positions.numpy() + faces = out_mesh.triangle.indices.numpy() + convex_vert_face_list.append((verts, faces)) + return extract_plane_equations(convex_vert_face_list) + + +def extract_plane_equations( + convex_meshes: List[Tuple[np.ndarray, np.ndarray]], +) -> List[Tuple[np.ndarray, np.ndarray]]: + """ + Extract plane equations from a list of convex hull meshes. + Each convex hull is represented by its outward-facing face normals and offsets. + No padding is applied; each hull can have a different number of faces. + + Args: + convex_meshes: List of convex hull data. + - tuple of (vertices [N,3], faces [M,3]) + + Returns: + List of (normals_i [Ki, 3], offsets_i [Ki]) tuples, one per convex hull. + Ki is the number of faces of the i-th hull and can differ across hulls. + """ + convex_plane_data = [] + + for i, convex_mesh_data in enumerate(convex_meshes): + vertices, faces = convex_mesh_data + hull = trimesh.Trimesh( + vertices=vertices, + faces=faces, + ) + # Outward-facing face normals [Ki, 3] + face_normals = hull.face_normals + # One vertex per face to compute offset [Ki, 3] + face_origins = hull.triangles[:, 0, :] + # Plane equation: n · x + d = 0 => d = -(n · p) + offsets_i = -np.sum(face_normals * face_origins, axis=1) + + convex_plane_data.append( + (face_normals.astype(np.float32), offsets_i.astype(np.float32)) + ) + return convex_plane_data + + +def sample_surface_points(mesh_path: str, num_points: int = 4096) -> np.ndarray: + """ + Sample surface points from a mesh file. + + Args: + mesh_path: Path to the mesh file. + num_points: Number of surface points to sample. + + Returns: + points: [P, 3] numpy array of sampled surface points. + """ + mesh = trimesh.load(mesh_path, force="mesh") + points = mesh.sample(num_points) + return points.astype(np.float32) + + +def check_collision_single_hull( + normals: torch.Tensor, # [K, 3] + offsets: torch.Tensor, # [K] + transformed_points: torch.Tensor, # [B, P, 3] + threshold: float = 0.0, +) -> Tuple[torch.Tensor, torch.Tensor]: + """ + Check collision between a batch of transformed point clouds and a single convex hull. + + A point p is inside the convex hull iff: + max_k (n_k · p + d_k) <= 0 + + Penetration depth for a point is defined as: + penetration = -(max_k (n_k · p + d_k)) + Positive penetration means the point is inside the hull. + + Args: + normals: [K, 3] outward face normals of the convex hull. + offsets: [K] plane offsets of the convex hull. + transformed_points: [B, P, 3] point cloud already transformed by batch poses. + threshold: collision threshold. A point is considered colliding if + its signed distance to the hull interior is <= threshold. + + Returns: + penetration: [B, P] penetration depth for each point. + Positive values indicate the point is inside the hull. + collides: [B, P] boolean mask, True if the point collides with this hull. + """ + # signed_dist: [B, P, K] = einsum([B,P,3], [K,3]) + [K] + signed_dist = torch.einsum("bpj, kj -> bpk", transformed_points, normals) + offsets + + # For each point, the maximum signed distance across all planes + # If max <= 0, the point satisfies all half-plane constraints => inside the hull + max_over_planes, _ = signed_dist.max(dim=-1) # [B, P] + + # Penetration depth: negate so that positive = inside + penetration = -max_over_planes # [B, P] + + # A point collides if its penetration exceeds the threshold + collides = penetration > threshold # [B, P] + + return penetration, collides diff --git a/embodichain/toolkits/graspkit/pg_grasp/cone_sampler.py b/embodichain/toolkits/graspkit/pg_grasp/cone_sampler.py deleted file mode 100644 index 7c9738fb..00000000 --- a/embodichain/toolkits/graspkit/pg_grasp/cone_sampler.py +++ /dev/null @@ -1,121 +0,0 @@ -# ---------------------------------------------------------------------------- -# Copyright (c) 2021-2026 DexForce Technology Co., Ltd. -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. -# ---------------------------------------------------------------------------- - -import open3d as o3d -import numpy as np -from scipy.spatial.transform import Rotation as R - - -def rotate_to_ref(direc: np.ndarray, rotate_ref: np.ndarray): - assert direc.shape == (3,) - direc_len = np.linalg.norm(direc) - assert direc_len > 1e-5 - direc_unit = direc / direc_len - - assert rotate_ref.shape == (3,) - rotate_ref_len = np.linalg.norm(rotate_ref) - assert rotate_ref_len > 1e-5 - rotate_ref_unit = rotate_ref / rotate_ref_len - - rotate_axis = np.cross(rotate_ref_unit, direc_unit) - rotate_axis_len = np.linalg.norm(rotate_axis) - if rotate_axis_len < 1e-5: - # co axis, no need to do rotation - dot_res = direc_unit.dot(rotate_ref_unit) - if dot_res > 0: - # identity rotation - return np.eye(3, dtype=float) - else: - # negative, rotate 180 degree - # rotate with a perpendicular axis - random_axis = np.random.random(size=(3,)) - perpendicular_axis = np.cross(random_axis, rotate_ref_unit) - perpendicular_axis = perpendicular_axis / np.linalg.norm(perpendicular_axis) - ref_rotation = R.from_rotvec(perpendicular_axis * np.pi).as_matrix() - return ref_rotation - else: - rotate_axis = rotate_axis / rotate_axis_len - angle = np.arccos(direc_unit.dot(rotate_ref_unit)) - ref_rotation = R.from_rotvec(angle * rotate_axis, degrees=False).as_matrix() - return ref_rotation - - -class ConeSampler: - def __init__( - self, max_angle: float, layer_num: int = 2, sample_each_layer: int = 4 - ) -> None: - """cone ray sampler - - Args: - max_angle (float): maximum ray angle to surface normal - layer_num (int, optional): circle layer. Defaults to 2. - sample_each_layer (int, optional): ray samples in each circle layer. Defaults to 4. - """ - self._max_angle = max_angle - self._layer_num = layer_num - self._ray_num = layer_num * sample_each_layer + 1 - alpha_list = np.linspace(max_angle / layer_num, max_angle, layer_num) - beta_list = np.linspace( - 2 * np.pi / sample_each_layer, 2 * np.pi, sample_each_layer - ) - self._direc_ref = np.array([0, 0, 1]) - - rotation_list = np.empty(shape=(self._ray_num, 3, 3), dtype=float) - - for i, alpha in enumerate(alpha_list): - for j, beta in enumerate(beta_list): - x_rotation = R.from_euler( - seq="XYZ", angles=np.array([alpha, 0, 0]), degrees=False - ).as_matrix() - z_rotation = R.from_euler( - seq="XYZ", angles=np.array([0, 0, beta]), degrees=False - ).as_matrix() - rotation_list[i * sample_each_layer + j + 1] = z_rotation @ x_rotation - # original direction - rotation_list[0] = np.eye(3) - self._sample_direc = rotation_list[:, :3, 2] # z-axis - - def cone_sample_direc(self, direc: np.ndarray, is_visual: bool = False): - """sample cone directly - - Args: - direc (np.ndarray): direction to sample a cone - is_visual (bool, optional): use visualization or not. Defaults to False. - - Returns: - np.ndarray: [_ray_num, 3] of float, cone direction list - """ - ref_rotation = rotate_to_ref(direc, self._direc_ref) - cone_direc_list = self._sample_direc @ ref_rotation.T - if is_visual: - self._visual(cone_direc_list) - return cone_direc_list - - def _visual(self, cone_direc_list: np.ndarray): - drawer = o3d.geometry.TriangleMesh.create_coordinate_frame(0.5) - for cone_direc in cone_direc_list: - arrow = o3d.geometry.TriangleMesh.create_arrow( - cylinder_radius=0.02, - cone_radius=0.03, - cylinder_height=0.9, - cone_height=0.1, - ) - arrow.compute_vertex_normals() - arrow_rotation = rotate_to_ref(cone_direc, self._direc_ref) - arrow.rotate(arrow_rotation, center=(0, 0, 0)) - arrow.paint_uniform_color(np.array([0.5, 0.5, 0.5])) - drawer += arrow - o3d.visualization.draw_geometries([drawer]) diff --git a/embodichain/toolkits/graspkit/pg_grasp/gripper_collision_checker.py b/embodichain/toolkits/graspkit/pg_grasp/gripper_collision_checker.py new file mode 100644 index 00000000..5f02176c --- /dev/null +++ b/embodichain/toolkits/graspkit/pg_grasp/gripper_collision_checker.py @@ -0,0 +1,266 @@ +# ---------------------------------------------------------------------------- +# Copyright (c) 2021-2026 DexForce Technology Co., Ltd. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# ---------------------------------------------------------------------------- + +from __future__ import annotations + +import torch + +from typing import Sequence + +from embodichain.utils import configclass +from embodichain.toolkits.graspkit.pg_grasp.collision_checker import ( + ConvexCollisionChecker, +) +from embodichain.utils.math import transform_points_mat + +__all__ = ["GripperCollisionCfg", "GripperCollisionChecker", "box_surface_grid"] + + +@configclass +class GripperCollisionCfg: + """Configuration for the GripperCollisionChecker. This class defines various parameters related to the + gripper geometry, point cloud generation, and collision checking process. Users can customize these parameters + based on the specific gripper being modeled and the requirements of the application. + """ + + max_open_length: float = 0.1 + """ Maximum opening length of the gripper fingers. This should be set according to the specific gripper being modeled, + and it defines the maximum distance between the two fingers when fully open. + """ + + finger_length: float = 0.08 + """ Length of the gripper fingers from the root to the tip, in z axis. This should be set according to the specific + gripper being modeled, and it defines how far the fingers extend from the gripper root frame. + """ + + y_thickness: float = 0.03 + """ Thickness of the gripper along the Y-axis (the axis perpendicular to the finger opening direction). This should + be set according to the specific gripper being modeled, and it defines the width of the gripper's main body and fingers + in the Y direction. + """ + + x_thickness: float = 0.01 + """ Thickness of the gripper along the X-axis (the axis parallel to the finger opening direction). This should + be set according to the specific gripper being modeled, and it defines the thickness of the fingers and the root + in the X direction. + """ + + root_z_width: float = 0.08 + """ Width of the gripper root along the Z-axis (the axis along the finger length direction). This should be set + according to the specific gripper being modeled, and it defines how far the root extends along the Z direction. + """ + + point_sample_dense: float = 0.01 + """ Approximate number of points per unit length for the gripper point cloud. Higher values will yield denser point + clouds, which can improve collision checking accuracy but also increase computational cost. This should be set based + on the desired balance between accuracy and efficiency for the specific application. + """ + + max_decomposition_hulls: int = 16 + """ Maximum number of convex hulls to decompose the object mesh into for collision checking. This should be set based + on the complexity of the object geometry and the desired accuracy of collision checking. More hulls can provide a tighter + approximation of the object shape but will increase computational cost. + """ + + open_check_margin: float = 0.01 + """ Additional margin added to the gripper open length when checking for collisions. This can help account for + uncertainties in the gripper pose or object geometry, and can be set based on the specific requirements of the application. + """ + + +class GripperCollisionChecker: + def __init__( + self, + object_mesh_verts: torch.Tensor, + object_mesh_faces: torch.Tensor, + cfg: GripperCollisionCfg = GripperCollisionCfg(), + ): + self._checker = ConvexCollisionChecker( + base_mesh_verts=object_mesh_verts, + base_mesh_faces=object_mesh_faces, + max_decomposition_hulls=cfg.max_decomposition_hulls, + ) + self.device = object_mesh_verts.device + self.cfg = cfg + self._init_pc_template() + + def _init_pc_template(self): + self.root_template = box_surface_grid( + size=( + self.cfg.max_open_length, + self.cfg.y_thickness, + self.cfg.root_z_width, + ), + dense=self.cfg.point_sample_dense, + device=self.device, + ) + self.left_template = box_surface_grid( + size=(self.cfg.x_thickness, self.cfg.y_thickness, self.cfg.finger_length), + dense=self.cfg.point_sample_dense, + device=self.device, + ) + self.right_template = box_surface_grid( + size=(self.cfg.x_thickness, self.cfg.y_thickness, self.cfg.finger_length), + dense=self.cfg.point_sample_dense, + device=self.device, + ) + + def _get_gripper_pc( + self, grasp_poses: torch.Tensor, open_lengths: torch.Tensor + ) -> torch.Tensor: + """ + Args: + grasp_poses: [B, 4, 4] homogeneous transformation matrix of the gripper root frame. + open_lengths: [B] opening length of the gripper fingers. + Returns: + gripper_pc: [B, P, 3] point cloud of the gripper in the world frame. + """ + + root_grasp_poses = grasp_poses.clone() + root_grasp_poses[:, :3, 3] -= ( + root_grasp_poses[:, :3, 2] + * 0.5 + * (self.cfg.finger_length + self.cfg.root_z_width) + ) + open_lengths_repeat = ( + open_lengths[:, None] + self.cfg.open_check_margin + ).repeat(1, 3) + left_finger_poses = grasp_poses.clone() + left_finger_poses[:, :3, 3] -= left_finger_poses[:, :3, 0] * open_lengths_repeat + + right_finger_poses = grasp_poses.clone() + right_finger_poses[:, :3, 3] += ( + right_finger_poses[:, :3, 0] * open_lengths_repeat + ) + + root_pc = transform_points_mat(self.root_template, root_grasp_poses) + left_pc = transform_points_mat(self.left_template, left_finger_poses) + right_pc = transform_points_mat(self.right_template, right_finger_poses) + gripper_pc = torch.cat([root_pc, left_pc, right_pc], dim=1) + return gripper_pc + + def query( + self, + obj_pose: torch.Tensor, + grasp_poses: torch.Tensor, + open_lengths: torch.Tensor, + collision_threshold: float = 0.0, + is_visual: bool = False, + ) -> torch.Tensor: + inv_obj_pose = obj_pose.clone() + inv_obj_pose[:3, :3] = obj_pose[:3, :3].T + inv_obj_pose[:3, 3] = -obj_pose[:3, 3] @ obj_pose[:3, :3] + inv_obj_poses = inv_obj_pose[None, :, :].repeat(grasp_poses.shape[0], 1, 1) + grasp_relative_pose = torch.bmm(inv_obj_poses, grasp_poses) + gripper_pc = self._get_gripper_pc(grasp_relative_pose, open_lengths) + return self._checker.query_batch_points( + gripper_pc, collision_threshold=collision_threshold, is_visual=is_visual + ) + + +def box_surface_grid( + size: Sequence[float] | torch.Tensor, + dense: float, + device: torch.device | str = "cpu", +) -> torch.Tensor: + """Generate grid-sampled points on the surface of an axis-aligned box. + + Six faces of the box are each sampled independently on a regular 2-D grid. + Grid resolution per face is derived automatically from ``dense``: + the number of sample points along an edge of length *L* is + ``max(2, round(L * dense) + 1)``, so ``dense`` behaves as + *approximate samples per unit length*. + + Edge and corner points are shared across adjacent faces and are included + exactly once (no duplicates). + + Args: + size: Box dimensions ``(sx, sy, sz)``. Accepts a sequence of three + floats or a 1-D :class:`torch.Tensor` of length 3. + dense: Approximate number of grid sample points per unit length along + each edge. Higher values yield denser point clouds. + device: Target PyTorch device for the returned tensor. + + Returns: + Float tensor of shape ``(N, 3)`` containing surface points expressed + in the box's local frame (origin at the box centre). + + Example: + >>> pts = box_surface_grid((0.1, 0.06, 0.03), dense=200.0) + >>> pts.shape + torch.Size([..., 3]) + """ + if isinstance(size, torch.Tensor): + sx, sy, sz = size[0].item(), size[1].item(), size[2].item() + else: + sx, sy, sz = float(size[0]), float(size[1]), float(size[2]) + + hx, hy, hz = sx / 2.0, sy / 2.0, sz / 2.0 + + # ── grid resolution per axis (at least 2 points to span the full edge) ── + nx = max(2, round(sx / dense) + 1) + ny = max(2, round(sy / dense) + 1) + nz = max(2, round(sz / dense) + 1) + + xs = torch.linspace(-hx, hx, nx, device=device) + ys = torch.linspace(-hy, hy, ny, device=device) + zs = torch.linspace(-hz, hz, nz, device=device) + + # Interior slices (exclude first and last to avoid duplicate edges) + xs_inner = xs[1:-1] # length nx-2 + ys_inner = ys[1:-1] # length ny-2 + + def _grid( + u: torch.Tensor, v: torch.Tensor, axis: int, offset: float + ) -> torch.Tensor: + """Build a flat (M, 3) tensor for one face grid. + + Args: + u: 1-D tensor of coordinates along the first in-plane axis. + v: 1-D tensor of coordinates along the second in-plane axis. + axis: Normal axis of the face — 0 (±X), 1 (±Y), or 2 (±Z). + offset: Signed half-extent along ``axis``. + + Returns: + Tensor of shape ``(len(u) * len(v), 3)``. + """ + uu, vv = torch.meshgrid(u, v, indexing="ij") + uu = uu.reshape(-1) + vv = vv.reshape(-1) + cc = torch.full_like(uu, offset) + if axis == 0: + return torch.stack([cc, uu, vv], dim=-1) + elif axis == 1: + return torch.stack([uu, cc, vv], dim=-1) + else: + return torch.stack([uu, vv, cc], dim=-1) + + # ───────────────────────────────────────────────────────────────────────── + # Build 6 faces. To avoid duplicate points on shared edges/corners: + # ±X faces → full NY × NZ grids + # ±Y faces → (NX-2) × NZ grids (x-edges owned by ±X faces) + # ±Z faces → (NX-2) × (NY-2) grids (x- and y-edges owned above) + # ───────────────────────────────────────────────────────────────────────── + faces: list[torch.Tensor] = [ + _grid(ys, zs, axis=0, offset=-hx), # −X face (NY × NZ) + _grid(ys, zs, axis=0, offset=+hx), # +X face (NY × NZ) + _grid(xs_inner, zs, axis=1, offset=-hy), # −Y face ((NX-2) × NZ) + _grid(xs_inner, zs, axis=1, offset=+hy), # +Y face ((NX-2) × NZ) + _grid(xs_inner, ys_inner, axis=2, offset=-hz), # −Z face + _grid(xs_inner, ys_inner, axis=2, offset=+hz), # +Z face + ] + + return torch.cat(faces, dim=0) diff --git a/embodichain/toolkits/graspkit/scripts/__init__.py b/embodichain/toolkits/graspkit/scripts/__init__.py new file mode 100644 index 00000000..dd650e90 --- /dev/null +++ b/embodichain/toolkits/graspkit/scripts/__init__.py @@ -0,0 +1,15 @@ +# ---------------------------------------------------------------------------- +# Copyright (c) 2021-2026 DexForce Technology Co., Ltd. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# ---------------------------------------------------------------------------- diff --git a/embodichain/toolkits/graspkit/scripts/annotate_grasp.py b/embodichain/toolkits/graspkit/scripts/annotate_grasp.py new file mode 100644 index 00000000..3e1c9211 --- /dev/null +++ b/embodichain/toolkits/graspkit/scripts/annotate_grasp.py @@ -0,0 +1,131 @@ +# ---------------------------------------------------------------------------- +# Copyright (c) 2021-2026 DexForce Technology Co., Ltd. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# ---------------------------------------------------------------------------- + +"""CLI for interactive grasp region annotation on a mesh. + +Loads a mesh file via *trimesh*, launches a browser-based annotator so the +user can select the graspable region, and saves the resulting antipodal +point pairs to the grasp-annotator cache. + +Usage examples:: + + python -m embodichain annotate-grasp --mesh_path /path/to/object.ply + python -m embodichain annotate-grasp --mesh_path mug.obj +""" + +from __future__ import annotations + +import argparse + +import torch +import trimesh + +from embodichain.toolkits.graspkit.pg_grasp import ( + AntipodalSamplerCfg, + GraspGenerator, + GraspGeneratorCfg, +) +from embodichain.utils.logger import log_info + + +def cli() -> None: + """Command-line interface for grasp pose annotation. + + Parses CLI arguments, loads the mesh, and launches interactive + annotation via the Viser browser UI. + """ + parser = argparse.ArgumentParser( + description=( + "Interactively annotate a grasp region on a mesh and " + "compute antipodal point pairs." + ), + ) + + parser.add_argument( + "--mesh_path", + type=str, + required=True, + help="Path to the mesh file (e.g. .ply, .obj, .stl).", + ) + parser.add_argument( + "--viser_port", + type=int, + default=15531, + help="Port for the browser-based annotation UI (default: 15531).", + ) + parser.add_argument( + "--n_sample", + type=int, + default=20000, + help="Number of surface points to sample (default: 20000).", + ) + parser.add_argument( + "--max_length", + type=float, + default=0.1, + help="Maximum distance between antipodal pairs in metres (default: 0.1).", + ) + parser.add_argument( + "--min_length", + type=float, + default=0.001, + help="Minimum distance between antipodal pairs in metres (default: 0.001).", + ) + parser.add_argument( + "--device", + type=str, + default="cpu", + help="Compute device, e.g. 'cpu' or 'cuda' (default: cpu).", + ) + + args = parser.parse_args() + + # Load mesh via trimesh + log_info(f"Loading mesh from {args.mesh_path}", color="green") + mesh = trimesh.load(args.mesh_path, force="mesh") + vertices = torch.tensor(mesh.vertices, dtype=torch.float32, device=args.device) + triangles = torch.tensor(mesh.faces, dtype=torch.int64, device=args.device) + + # Build configuration + sampler_cfg = AntipodalSamplerCfg( + n_sample=args.n_sample, + max_length=args.max_length, + min_length=args.min_length, + ) + cfg = GraspGeneratorCfg( + viser_port=args.viser_port, + antipodal_sampler_cfg=sampler_cfg, + ) + + # Create generator and run annotation + generator = GraspGenerator(vertices=vertices, triangles=triangles, cfg=cfg) + log_info( + "Annotate the grasp region in the browser window:\n" + " 1. Open http://localhost:{port}\n" + " 2. Click 'Rect Select Region' and drag to select\n" + " 3. Click 'Confirm Selection' to finish", + color="green", + ) + hit_point_pairs = generator.annotate() + + log_info( + f"Annotation complete. {hit_point_pairs.shape[0]} antipodal pairs cached.", + color="green", + ) + + +if __name__ == "__main__": + cli() diff --git a/embodichain/utils/math.py b/embodichain/utils/math.py index 084e51f4..caaa39d2 100644 --- a/embodichain/utils/math.py +++ b/embodichain/utils/math.py @@ -1206,6 +1206,25 @@ def transform_points( return points_batch +def transform_points_mat( + points: torch.Tensor, poses: torch.Tensor # [P, 3] # [B, 4, 4] +) -> torch.Tensor: + """ + Apply a batch of rigid transforms to a point cloud. + + Args: + points: [P, 3] source point cloud. + poses: [B, 4, 4] batch of homogeneous transformation matrices. + + Returns: + transformed: [B, P, 3] transformed point cloud for each pose. + """ + R = poses[:, :3, :3] # [B, 3, 3] + t = poses[:, :3, 3] # [B, 3] + transformed = torch.einsum("bij, pj -> bpi", R, points) + t.unsqueeze(1) + return transformed + + """ Projection operations. """ diff --git a/embodichain/utils/warp/__init__.py b/embodichain/utils/warp/__init__.py index 905bc9e7..c08be1d5 100644 --- a/embodichain/utils/warp/__init__.py +++ b/embodichain/utils/warp/__init__.py @@ -30,3 +30,5 @@ repeat_first_point, interpolate_along_distance, ) + +from .collision.convex_query import convex_signed_distance_kernel diff --git a/embodichain/utils/warp/collision/__init__.py b/embodichain/utils/warp/collision/__init__.py new file mode 100644 index 00000000..d7e19801 --- /dev/null +++ b/embodichain/utils/warp/collision/__init__.py @@ -0,0 +1,17 @@ +# ---------------------------------------------------------------------------- +# Copyright (c) 2021-2026 DexForce Technology Co., Ltd. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# ---------------------------------------------------------------------------- + +from . import convex_query diff --git a/embodichain/utils/warp/collision/convex_query.py b/embodichain/utils/warp/collision/convex_query.py new file mode 100644 index 00000000..ce2e7c1f --- /dev/null +++ b/embodichain/utils/warp/collision/convex_query.py @@ -0,0 +1,54 @@ +# ---------------------------------------------------------------------------- +# Copyright (c) 2021-2026 DexForce Technology Co., Ltd. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# ---------------------------------------------------------------------------- + +import warp as wp + + +@wp.kernel(enable_backward=False) +def convex_signed_distance_kernel( + query_points: wp.array(dtype=wp.float32, ndim=3), + plane_equations: wp.array(dtype=wp.float32, ndim=3), + plane_equation_counts: wp.array(dtype=wp.int32, ndim=1), + signed_distances: wp.array(dtype=wp.float32, ndim=3), +): + """ + Compute the signed distance from query points to convex hulls defined by plane equations. + + Args: + query_points: [n_pose, n_point, 3] coordinates of query points. + plane_equations: [n_convex, n_max_equation, 4] plane equations of convex hulls, where each plane equation is represented as (normal_x, normal_y, normal_z, offset). + plane_equation_counts: [n_convex, ] number of valid plane equations for each convex hull. + + Returns: + signed_distances: [n_pose, n_point, n_convex] output signed distances from query points to convex hulls. Should be initialized as +inf before calling this kernel. + """ + pose_id, point_id, convex_id = wp.tid() + n_equation = plane_equation_counts[convex_id] + for i in range(n_equation): + normal_x = plane_equations[convex_id, i, 0] + normal_y = plane_equations[convex_id, i, 1] + normal_z = plane_equations[convex_id, i, 2] + offset = plane_equations[convex_id, i, 3] + signed_distance = ( + query_points[pose_id, point_id, 0] * normal_x + + query_points[pose_id, point_id, 1] * normal_y + + query_points[pose_id, point_id, 2] * normal_z + + offset + ) + # should initialize as -inf + signed_distances[pose_id, point_id, convex_id] = max( + signed_distance, signed_distances[pose_id, point_id, convex_id] + ) diff --git a/pyproject.toml b/pyproject.toml index 60a12496..25b15290 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -50,7 +50,8 @@ dependencies = [ "black==24.3.0", "fvcore", "h5py", - "tensordict" + "tensordict", + "viser==1.0.21" ] [project.optional-dependencies] diff --git a/scripts/tutorials/grasp/grasp_generator.py b/scripts/tutorials/grasp/grasp_generator.py new file mode 100644 index 00000000..bab09c03 --- /dev/null +++ b/scripts/tutorials/grasp/grasp_generator.py @@ -0,0 +1,290 @@ +# ---------------------------------------------------------------------------- +# Copyright (c) 2021-2026 DexForce Technology Co., Ltd. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# ---------------------------------------------------------------------------- + +""" +This script demonstrates the creation and simulation of a robot that grasps a rigid mug +in a simulated environment using the SimulationManager and grasp planning utilities. +""" + +import argparse +import numpy as np +import time +import torch + +from embodichain.lab.sim import SimulationManager, SimulationManagerCfg +from embodichain.lab.sim.objects import Robot, RigidObject +from embodichain.lab.sim.utility.action_utils import interpolate_with_distance +from embodichain.lab.sim.shapes import MeshCfg +from embodichain.lab.sim.solvers import PytorchSolverCfg +from embodichain.data import get_data_path +from embodichain.utils import logger +from embodichain.lab.sim.cfg import ( + JointDrivePropertiesCfg, + RobotCfg, + LightCfg, + RigidBodyAttributesCfg, + RigidObjectCfg, + URDFCfg, +) +from embodichain.toolkits.graspkit.pg_grasp.antipodal_generator import ( + GraspGenerator, + GraspGeneratorCfg, + AntipodalSamplerCfg, +) +from embodichain.toolkits.graspkit.pg_grasp.gripper_collision_checker import ( + GripperCollisionCfg, +) + + +def parse_arguments(): + """ + Parse command-line arguments to configure the simulation. + + Returns: + argparse.Namespace: Parsed arguments including number of environments and rendering options. + """ + parser = argparse.ArgumentParser( + description="Create and simulate a robot in SimulationManager" + ) + parser.add_argument( + "--num_envs", type=int, default=1, help="Number of parallel environments" + ) + parser.add_argument( + "--enable_rt", action="store_true", help="Enable ray tracing rendering" + ) + parser.add_argument("--headless", action="store_true", help="Enable headless mode") + parser.add_argument( + "--device", + type=str, + default="cpu", + help="device to run the environment on, e.g., 'cpu' or 'cuda'", + ) + return parser.parse_args() + + +def initialize_simulation(args) -> SimulationManager: + """ + Initialize the simulation environment based on the provided arguments. + + Args: + args (argparse.Namespace): Parsed command-line arguments. + + Returns: + SimulationManager: Configured simulation manager instance. + """ + config = SimulationManagerCfg( + headless=True, + sim_device=args.device, + enable_rt=args.enable_rt, + physics_dt=1.0 / 100.0, + arena_space=2.5, + ) + sim = SimulationManager(config) + + if args.enable_rt: + light = sim.add_light( + cfg=LightCfg( + uid="main_light", + color=(0.6, 0.6, 0.6), + intensity=30.0, + init_pos=(1.0, 0, 3.0), + ) + ) + + return sim + + +def create_robot(sim: SimulationManager, position=[0.0, 0.0, 0.0]) -> Robot: + """ + Create and configure a robot with an arm and a dexterous hand in the simulation. + + Args: + sim (SimulationManager): The simulation manager instance. + + Returns: + Robot: The configured robot instance added to the simulation. + """ + # Retrieve URDF paths for the robot arm and hand + ur10_urdf_path = get_data_path("UniversalRobots/UR10/UR10.urdf") + gripper_urdf_path = get_data_path("DH_PGC_140_50_M/DH_PGC_140_50_M.urdf") + # Configure the robot with its components and control properties + cfg = RobotCfg( + uid="UR10", + urdf_cfg=URDFCfg( + components=[ + {"component_type": "arm", "urdf_path": ur10_urdf_path}, + {"component_type": "hand", "urdf_path": gripper_urdf_path}, + ] + ), + drive_pros=JointDrivePropertiesCfg( + stiffness={"JOINT[0-9]": 1e4, "FINGER[1-2]": 1e3}, + damping={"JOINT[0-9]": 1e3, "FINGER[1-2]": 1e2}, + max_effort={"JOINT[0-9]": 1e5, "FINGER[1-2]": 1e4}, + drive_type="force", + ), + control_parts={ + "arm": ["JOINT[0-9]"], + "hand": ["FINGER[1-2]"], + }, + solver_cfg={ + "arm": PytorchSolverCfg( + end_link_name="ee_link", + root_link_name="base_link", + tcp=[ + [0.0, 1.0, 0.0, 0.0], + [-1.0, 0.0, 0.0, 0.0], + [0.0, 0.0, 1.0, 0.12], + [0.0, 0.0, 0.0, 1.0], + ], + ) + }, + init_qpos=[0.0, -np.pi / 2, -np.pi / 2, np.pi / 2, -np.pi / 2, 0.0, 0.0, 0.0], + init_pos=position, + ) + return sim.add_robot(cfg=cfg) + + +def create_mug(sim: SimulationManager): + mug_cfg = RigidObjectCfg( + uid="table", + shape=MeshCfg( + fpath=get_data_path("CoffeeCup/cup.ply"), + ), + attrs=RigidBodyAttributesCfg( + mass=0.01, + dynamic_friction=0.97, + static_friction=0.99, + ), + max_convex_hull_num=16, + init_pos=[0.55, 0.0, 0.01], + init_rot=[0.0, 0.0, -90], + body_scale=(4, 4, 4), + ) + mug = sim.add_rigid_object(cfg=mug_cfg) + return mug + + +def get_grasp_traj(sim: SimulationManager, robot: Robot, grasp_xpos: torch.Tensor): + n_envs = sim.num_envs + rest_arm_qpos = robot.get_qpos("arm") + + approach_xpos = grasp_xpos.clone() + approach_xpos[:, 2, 3] += 0.1 + + _, qpos_approach = robot.compute_ik( + pose=approach_xpos, joint_seed=rest_arm_qpos, name="arm" + ) + _, qpos_grasp = robot.compute_ik( + pose=grasp_xpos, joint_seed=qpos_approach, name="arm" + ) + hand_open_qpos = torch.tensor([0.00, 0.00], dtype=torch.float32, device=sim.device) + hand_close_qpos = torch.tensor( + [0.025, 0.025], dtype=torch.float32, device=sim.device + ) + + arm_trajectory = torch.cat( + [ + rest_arm_qpos[:, None, :], + qpos_approach[:, None, :], + qpos_grasp[:, None, :], + qpos_grasp[:, None, :], + qpos_approach[:, None, :], + rest_arm_qpos[:, None, :], + ], + dim=1, + ) + hand_trajectory = torch.cat( + [ + hand_open_qpos[None, None, :].repeat(n_envs, 1, 1), + hand_open_qpos[None, None, :].repeat(n_envs, 1, 1), + hand_open_qpos[None, None, :].repeat(n_envs, 1, 1), + hand_close_qpos[None, None, :].repeat(n_envs, 1, 1), + hand_close_qpos[None, None, :].repeat(n_envs, 1, 1), + hand_close_qpos[None, None, :].repeat(n_envs, 1, 1), + ], + dim=1, + ) + all_trajectory = torch.cat([arm_trajectory, hand_trajectory], dim=-1) + interp_trajectory = interpolate_with_distance( + trajectory=all_trajectory, interp_num=200, device=sim.device + ) + return interp_trajectory + + +if __name__ == "__main__": + import time + + args = parse_arguments() + sim = initialize_simulation(args) + robot = create_robot(sim, position=[0.0, 0.0, 0.0]) + mug = create_mug(sim) + + # get mug grasp pose + grasp_cfg = GraspGeneratorCfg( + viser_port=11801, + antipodal_sampler_cfg=AntipodalSamplerCfg( + n_sample=20000, max_length=0.088, min_length=0.003 + ), + ) + sim.open_window() + + # Annotate part of the mug to be grasped by following the instructions in the visualization window: + # 1. View grasp object in browser (e.g http://localhost:11801) + # 2. press 'Rect Select Region', select grasp region + # 3. press 'Confirm Selection' to finish grasp region selection. + + start_time = time.time() + + gripper_collision_cfg = GripperCollisionCfg( + max_open_length=0.088, finger_length=0.078, point_sample_dense=0.012 + ) + + # Extract mesh data from the mug and create grasp generator + vertices = mug.get_vertices(env_ids=[0], scale=True)[0] + triangles = mug.get_triangles(env_ids=[0])[0] + grasp_generator = GraspGenerator( + vertices=vertices, + triangles=triangles, + cfg=grasp_cfg, + gripper_collision_cfg=gripper_collision_cfg, + ) + + # Annotate grasp region (populates internal antipodal point pairs) + grasp_generator.annotate() + + # Compute grasp poses per environment + approach_direction = torch.tensor( + [0, 0, -1], dtype=torch.float32, device=sim.device + ) + obj_poses = mug.get_local_pose(to_matrix=True) + grasp_xpos_list = [] + for obj_pose in obj_poses: + grasp_pose, _ = grasp_generator.get_grasp_poses( + obj_pose, approach_direction, visualize_pose=False + ) + grasp_xpos_list.append(grasp_pose.unsqueeze(0)) + grasp_xpos = torch.cat(grasp_xpos_list, dim=0) + cost_time = time.time() - start_time + logger.log_info(f"Get grasp pose cost time: {cost_time:.2f} seconds") + + grab_traj = get_grasp_traj(sim, robot, grasp_xpos) + input("Press Enter to start the grab mug demo...") + n_waypoint = grab_traj.shape[1] + for i in range(n_waypoint): + robot.set_qpos(grab_traj[:, i, :]) + sim.update(step=4) + time.sleep(1e-2) + input("Press Enter to exit the simulation...") diff --git a/tests/toolkits/test_batch_convex_collision.py b/tests/toolkits/test_batch_convex_collision.py new file mode 100644 index 00000000..4bf852c8 --- /dev/null +++ b/tests/toolkits/test_batch_convex_collision.py @@ -0,0 +1,77 @@ +# ---------------------------------------------------------------------------- +# Copyright (c) 2021-2026 DexForce Technology Co., Ltd. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# ---------------------------------------------------------------------------- +import torch +from embodichain.data import get_data_path +import trimesh +from embodichain.toolkits.graspkit.pg_grasp.collision_checker import ( + ConvexCollisionChecker, + ConvexCollisionCheckerCfg, +) +from embodichain.utils.math import transform_points_mat +import warp as wp + + +def batch_convex_collision_query(device=torch.device("cuda")): + mug_path = get_data_path("ScannedBottle/moliwulong_processed.ply") + mug_mesh = trimesh.load(mug_path, force="mesh", process=False) + verts = torch.tensor(mug_mesh.vertices, dtype=torch.float32, device=device) + faces = torch.tensor(mug_mesh.faces, dtype=torch.int32, device=device) + collision_checker = ConvexCollisionChecker(verts, faces, max_decomposition_hulls=16) + + poses = torch.tensor( + [ + [ + [1, 0, 0, 0], + [0, 1, 0, 0], + [0, 0, 1, 1.05], + [0, 0, 0, 1], + ], + [ + [1, 0, 0, 0.05], + [0, -1, 0, 0], + [0, 0, -1, 0], + [0, 0, 0, 1], + ], + ], + device=device, + ) + from scipy.spatial.transform import Rotation + + rot = Rotation.from_euler("xyz", [12, 3, 32], degrees=True).as_matrix() + poses[0, :3, :3] = torch.tensor(rot, dtype=torch.float32, device=device) + poses[1, :3, :3] = torch.tensor(rot, dtype=torch.float32, device=device) + + obj_path = get_data_path("ScannedBottle/yibao_processed.ply") + obj_mesh = trimesh.load(obj_path, force="mesh", process=False) + obj_verts = torch.tensor(obj_mesh.vertices, dtype=torch.float32, device=device) + obj_faces = torch.tensor(obj_mesh.faces, dtype=torch.int32, device=device) + test_pc = transform_points_mat(obj_verts, poses) + + is_pose_collide, pose_surface_distance = collision_checker.query_batch_points( + test_pc, collision_threshold=0.003, is_visual=False + ) + assert is_pose_collide.sum().item() == 1 + assert abs(pose_surface_distance.max().item() - 0.8492) < 1e-2 + + +def test_batch_convex_collision_cpu(): + wp.init() + batch_convex_collision_query(torch.device("cpu")) + + +def test_batch_convex_collision_gpu(): + wp.init() + batch_convex_collision_query(torch.device("cuda")) diff --git a/tests/toolkits/test_pg_grasp.py b/tests/toolkits/test_pg_grasp.py deleted file mode 100644 index 10f96bd7..00000000 --- a/tests/toolkits/test_pg_grasp.py +++ /dev/null @@ -1,96 +0,0 @@ -# ---------------------------------------------------------------------------- -# Copyright (c) 2021-2026 DexForce Technology Co., Ltd. -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. -# ---------------------------------------------------------------------------- - -import open3d as o3d -import numpy as np -import os -from embodichain.toolkits.graspkit.pg_grasp import ( - AntipodalGenerator, - GraspSelectMethod, -) -from embodichain.data import get_data_path - - -def test_antipodal_score_selector(is_visual: bool = False): - mesh_path = get_data_path("ChainRainSec/mesh.ply") - mesh_o3dt = o3d.t.io.read_triangle_mesh(mesh_path) - generator = AntipodalGenerator( - mesh_o3dt=mesh_o3dt, - open_length=0.1, - max_angle=np.pi / 6, - surface_sample_num=5000, - cache_dir=None, - ) - grasp_list = generator.select_grasp( - approach_direction=np.array([0, 0, -1]), - select_num=5, - select_method=GraspSelectMethod.NORMAL_SCORE, - ) - assert len(grasp_list) == 5 - if is_visual: - visual_mesh_list = generator.grasp_pose_visual(grasp_list) - visual_mesh_list = [visual_mesh.to_legacy() for visual_mesh in visual_mesh_list] - o3d.visualization.draw_geometries(visual_mesh_list) - - -def test_antipodal_position_selector(is_visual: bool = False): - mesh_path = get_data_path("ChainRainSec/mesh.ply") - mesh_o3dt = o3d.t.io.read_triangle_mesh(mesh_path) - generator = AntipodalGenerator( - mesh_o3dt=mesh_o3dt, - open_length=0.1, - max_angle=np.pi / 6, - surface_sample_num=5000, - cache_dir=None, - ) - grasp_list = generator.select_grasp( - approach_direction=np.array([0, 0, -1]), - select_num=5, - select_method=GraspSelectMethod.NEAR_APPROACH, - ) - assert len(grasp_list) == 5 - if is_visual: - visual_mesh_list = generator.grasp_pose_visual(grasp_list) - visual_mesh_list = [visual_mesh.to_legacy() for visual_mesh in visual_mesh_list] - o3d.visualization.draw_geometries(visual_mesh_list) - - -def test_antipodal_center_selector(is_visual: bool = False): - mesh_path = get_data_path("ChainRainSec/mesh.ply") - mesh_o3dt = o3d.t.io.read_triangle_mesh(mesh_path) - generator = AntipodalGenerator( - mesh_o3dt=mesh_o3dt, - open_length=0.1, - max_angle=np.pi / 6, - surface_sample_num=5000, - cache_dir=None, - ) - grasp_list = generator.select_grasp( - approach_direction=np.array([0, 0, -1]), - select_num=5, - select_method=GraspSelectMethod.CENTER, - ) - assert len(grasp_list) == 5 - if is_visual: - visual_mesh_list = generator.grasp_pose_visual(grasp_list) - visual_mesh_list = [visual_mesh.to_legacy() for visual_mesh in visual_mesh_list] - o3d.visualization.draw_geometries(visual_mesh_list) - - -if __name__ == "__main__": - test_antipodal_score_selector(True) - test_antipodal_position_selector(True) - test_antipodal_center_selector(True)