| import os, sys, math |
| import json, time |
| import numpy as np |
| import argparse |
| from isaacgym import gymtorch, gymapi |
| import torch |
|
|
| CURRENT_DIR = os.path.dirname(os.path.abspath(__file__)) |
| CSV_FRAME_RATE = 50.0 |
| CSV_TOTAL_COLS = 36 |
| CSV_ROOT_POS_SLICE = slice(0, 3) |
| CSV_ROOT_QUAT_SLICE = slice(3, 7) |
| CSV_DOF_SLICE = slice(7, 36) |
|
|
| def get_label_indices(labels, prefix): |
| return [i for i, label in enumerate(labels) if label.startswith(prefix)] |
|
|
| def load_motion_data_json(json_path): |
| with open(json_path, 'r') as f: |
| motion = json.load(f) |
|
|
| labels = motion.get('Labels', []) |
| frames = np.array(motion['Frames'], dtype=np.float32) |
|
|
| |
| dof_indices = get_label_indices(labels, "dof_pos/") |
| root_pos_indices = get_label_indices(labels, "root_pos/") |
| root_quat_indices = get_label_indices(labels, "root_quat/") |
|
|
| print(f"Found {len(dof_indices)} DOF channels:") |
| for idx in dof_indices: |
| print(f" {labels[idx]}") |
| print(f"root_pos indices: {[labels[i] for i in root_pos_indices]}") |
| print(f"root_quat indices: {[labels[i] for i in root_quat_indices]}") |
|
|
| return ( |
| frames[:, dof_indices], |
| frames[:, root_pos_indices], |
| frames[:, root_quat_indices], |
| motion['FrameDuration'] |
| ) |
|
|
| def load_motion_data_csv(csv_path): |
| frames = np.loadtxt(csv_path, delimiter=',', dtype=np.float32) |
| if frames.ndim == 1: |
| frames = frames[None, :] |
|
|
| if frames.shape[1] < CSV_TOTAL_COLS: |
| raise ValueError( |
| f"CSV requires at least {CSV_TOTAL_COLS} columns, got {frames.shape[1]}: {csv_path}" |
| ) |
|
|
| if frames.shape[1] > CSV_TOTAL_COLS: |
| print( |
| f"CSV has {frames.shape[1]} columns, using first {CSV_TOTAL_COLS} columns: {csv_path}" |
| ) |
| frames = frames[:, :CSV_TOTAL_COLS] |
|
|
| print(f"Loaded CSV motion: {csv_path}") |
| print(f"root_pos columns: {CSV_ROOT_POS_SLICE.start + 1}~{CSV_ROOT_POS_SLICE.stop}") |
| print(f"root_quat columns: {CSV_ROOT_QUAT_SLICE.start + 1}~{CSV_ROOT_QUAT_SLICE.stop}") |
| print(f"dof columns: {CSV_DOF_SLICE.start + 1}~{CSV_DOF_SLICE.stop}") |
|
|
| return ( |
| frames[:, CSV_DOF_SLICE], |
| frames[:, CSV_ROOT_POS_SLICE], |
| frames[:, CSV_ROOT_QUAT_SLICE], |
| 1.0 / CSV_FRAME_RATE |
| ) |
|
|
| def load_motion_data(motion_path): |
| ext = os.path.splitext(motion_path)[1].lower() |
| if ext == ".json": |
| return load_motion_data_json(motion_path) |
| if ext == ".csv": |
| return load_motion_data_csv(motion_path) |
| raise ValueError(f"Unsupported motion file format: {motion_path}") |
|
|
| def set_base_pose(gym, sim, env, actor, root_pos, root_quat): |
| """ |
| 设置机器人 base link 的位置和旋转 |
| :param root_pos: [x, y, z] |
| :param root_quat: [x, y, z, w] |
| """ |
| root_state_tensor = gym.acquire_actor_root_state_tensor(sim) |
| root_state = gymtorch.wrap_tensor(root_state_tensor) |
|
|
| |
| actor_index = 0 |
|
|
| |
| root_state[actor_index, :3] = torch.tensor(root_pos, dtype=torch.float32) |
| root_state[actor_index, 3:7] = torch.tensor(root_quat, dtype=torch.float32) |
|
|
| |
| gym.set_actor_root_state_tensor(sim, root_state_tensor) |
|
|
| def set_dof_positions(gym, env, actor, target_pos): |
| dof_states = np.zeros(len(target_pos), dtype=gymapi.DofState.dtype) |
| dof_states['pos'] = target_pos |
| dof_states['vel'] = np.zeros_like(target_pos) |
| gym.set_actor_dof_states(env, actor, dof_states, gymapi.STATE_POS + gymapi.STATE_VEL) |
|
|
| def parse_args(): |
| parser = argparse.ArgumentParser(description="Isaac Gym Preview 4 Motion Visualizer") |
|
|
| parser.add_argument('--file_name', type=str, |
| default="mixamo/low_jump.json", |
| help='Motion file name (.json or .csv). Default: "mixamo/low_jump.json"') |
|
|
| parser.add_argument('--robot_type', type=str, |
| choices=['adam_lite', 'adam_sp'], |
| default='adam_lite', |
| help='Robot type to load. Default: "adam_lite"') |
|
|
| return parser.parse_args() |
|
|
| def resolve_existing_path(candidates): |
| for rel_path in candidates: |
| abs_path = os.path.join(CURRENT_DIR, rel_path) |
| if os.path.exists(abs_path): |
| return rel_path |
| raise FileNotFoundError(f"No valid path found in candidates: {candidates}") |
|
|
| def get_robot_paths(robot_type): |
| if robot_type == "adam_lite": |
| urdf_path = resolve_existing_path([ |
| "robot_description/adam_lite/urdf/adam_lite.urdf", |
| "robot_description/adam_lite/adam_lite.urdf", |
| ]) |
| motion_dir = "adam_lite" |
| elif robot_type == "adam_sp": |
| urdf_path = resolve_existing_path([ |
| "robot_description/adam_sp/urdf/adam_sp.urdf", |
| "robot_description/adam_sp/adam_sp.urdf", |
| ]) |
| motion_dir = "adam_sp" |
| else: |
| raise ValueError(f"Unknown robot_type: {robot_type}") |
|
|
| return urdf_path, motion_dir |
|
|
| def main(): |
| args = parse_args() |
| |
| asset_path, motion_dir = get_robot_paths(args.robot_type) |
| asset_root = os.path.dirname(asset_path) |
| asset_file = os.path.basename(asset_path) |
|
|
| |
| motion_path = os.path.join(CURRENT_DIR, motion_dir, args.file_name) |
| frames_dof, frames_root_pos, frames_root_quat, frame_duration = load_motion_data(motion_path) |
| num_frames = len(frames_dof) |
| current_frame = 0 |
| |
| gym = gymapi.acquire_gym() |
|
|
| |
| sim_params = gymapi.SimParams() |
| sim_params.dt = frame_duration |
| sim_params.substeps = 2 |
| sim_params.up_axis = gymapi.UP_AXIS_Z |
| sim_params.gravity = gymapi.Vec3(0.0, 0.0, 0.0) |
| sim_params.use_gpu_pipeline = False |
| sim = gym.create_sim(0, 0, gymapi.SIM_PHYSX, sim_params) |
|
|
| if sim is None: |
| raise Exception("Failed to create sim") |
|
|
| |
| plane_params = gymapi.PlaneParams() |
| plane_params.normal = gymapi.Vec3(0, 0, 1) |
| gym.add_ground(sim, plane_params) |
|
|
|
|
| asset_options = gymapi.AssetOptions() |
| asset_options.fix_base_link = False |
| asset_options.disable_gravity = True |
| asset_options.collapse_fixed_joints = True |
|
|
| robot_asset = gym.load_asset(sim, asset_root, asset_file, asset_options) |
|
|
| |
| num_dof = gym.get_asset_dof_count(robot_asset) |
| if frames_dof.shape[1] != num_dof: |
| print( |
| f"DOF count mismatch: motion has {frames_dof.shape[1]}, " |
| f"asset requires {num_dof}. Auto-aligning." |
| ) |
| if frames_dof.shape[1] > num_dof: |
| frames_dof = frames_dof[:, :num_dof] |
| else: |
| pad = np.zeros((frames_dof.shape[0], num_dof - frames_dof.shape[1]), dtype=np.float32) |
| frames_dof = np.concatenate([frames_dof, pad], axis=1) |
|
|
| |
| spacing = 1.0 |
| lower = gymapi.Vec3(-spacing, -spacing, 0.0) |
| upper = gymapi.Vec3(spacing, spacing, spacing) |
| env = gym.create_env(sim, lower, upper, 8) |
|
|
| |
| pose = gymapi.Transform() |
| pose.p = gymapi.Vec3(0.0, 0.0, 1.0) |
| pose.r = gymapi.Quat(0.0, 0.0, 0.0, 1.0) |
|
|
| actor = gym.create_actor(env, robot_asset, pose, "actor", 0, 0) |
|
|
| cam_props = gymapi.CameraProperties() |
| viewer = gym.create_viewer(sim, cam_props) |
| if viewer is None: |
| raise Exception("Failed to create viewer") |
|
|
| |
| CAMERA_DISTANCE = -3.0 |
| CAMERA_HEIGHT = 1.0 |
| CAMERA_ANGLE = math.radians(90) |
| while not gym.query_viewer_has_closed(viewer): |
| if current_frame < num_frames: |
| target_dof = frames_dof[current_frame] |
| target_root_pos = frames_root_pos[current_frame] |
| target_root_quat = frames_root_quat[current_frame] |
|
|
| |
| set_base_pose(gym, sim, env, actor, target_root_pos, target_root_quat) |
|
|
| |
| set_dof_positions(gym, env, actor, target_dof) |
|
|
| current_frame += 1 |
| else: |
| print("Animation finished.") |
| break |
|
|
| |
| gym.simulate(sim) |
| gym.fetch_results(sim, True) |
| gym.step_graphics(sim) |
| |
| root_state_tensor = gym.acquire_actor_root_state_tensor(sim) |
| root_state = gymtorch.wrap_tensor(root_state_tensor) |
| root_pos = root_state[0, :3].cpu().numpy() |
|
|
| |
| cam_x = root_pos[0] + CAMERA_DISTANCE * math.cos(CAMERA_ANGLE) |
| cam_y = root_pos[1] + CAMERA_DISTANCE * math.sin(CAMERA_ANGLE) |
| cam_z = root_pos[2] + CAMERA_HEIGHT |
|
|
| cam_pos = gymapi.Vec3(cam_x, cam_y, cam_z) |
| cam_target = gymapi.Vec3(root_pos[0], root_pos[1], root_pos[2]) |
|
|
| |
| gym.viewer_camera_look_at(viewer, None, cam_pos, cam_target) |
|
|
| gym.draw_viewer(viewer, sim, True) |
| gym.sync_frame_time(sim) |
| time.sleep(0.001) |
|
|
| if __name__ == "__main__": |
| main() |
|
|