Project structure of: Genesis-Embodied-AI/RoboGen

• RoboGen Advanced AI Tools and Utilities
  • cem_policy Parallel cost calculation and gif saving utilities.
    • parallel_worker.py Parallel multiprocessing cost calculation.
    • utils.py Saves numpy array images as gif with environment settings.
  • environment.yaml RoboGen/environment.yaml - Python dependencies for RoboGen environment and data science projects
  • execute.py Execute.py: Prepares, tests, executes, saves state for RoboGen RL algorithms.
  • execute_locomotion.py Optimize cost functions, set up RL in assistive gym envs.
  • execute_long_horizon.py Executes tasks, writes JSON meta, detects GIFs, saves results, selects best outcome.
  • gpt_4 GPT-4 resources, functions, APIs for text processing.
    • adjust_size.py Adjust 3D model size based on object type from YAML input.
    • bard_verify.py Image verification with Bard API.
    • prompts Robotics AI tasks, configurations, URDF, parsing for RoboGen.
      • prompt_distractor.py Distractor config generation for Partnet Mobility task.
      • prompt_from_description.py Robot arm, tasks, articulation tree, grippers, config paths, distractor files.
      • prompt_locomotion.py Reinforcement learning, GPT tasks, YAML configs for quadruped/humanoid locomotion.
      • prompt_manipulation.py Franka Panda arm household manipulation tasks generator.
      • prompt_manipulation_reward_primitive.py Robotic household task completion using primitives, rewards, and Gym.
      • prompt_set_joint_angle.py Calculates joint angles for articulated objects and robots.
      • prompt_spatial_relationship.py Identifies spatial relationships between furniture and drawers in refrigerator.
      • prompt_with_scale.py Robotic arm simulator scales objects realistically with NLP.
      • utils.py YAML configurations, URDF refrigerators, parsing for RoboGen tasks.
    • query.py Imports openai, defines query function with debugging & API usage.
    • verification.py Verifies text similarity using GPU-based embeddings
  • install_ompl_1.5.2.sh Install ROS dependencies, libraries, Python packages; optionally install OMPL on Ubuntu.
  • locomotion Locomotion Meshes for Reward Calculation
    • atlas Convert .dae to .obj using Assimp for atlas locomotion meshes
      • meshes_unplugged
        • convert.bat Convert robot mesh files from .dae to .obj format.
      • meshes_v3
        • convert.bat Converts DAE to OBJ with Assimp
      • multisense_sl_description
        • meshes
          • convert.bat Convert .dae to .obj using Assimp
    • gpt_reward_api.py Calculates energy reward for controlling robotic movement from PyBullet simulator.
    • sim.py Robot simulation setup with PyBullet, observation and action spaces.
  • manipulation Robotics manipulation and 3D simulation.
    • agent.py Agent: Robot joint, physics engine, state management for manipulation.
    • assets
      • franka_mobile
        • LICENSE.txt Apache License 2.0: Use, Modify, Distribute, Liability Limitation
    • gpt_primitive_api.py Grasping primitive API for robot manipulation.
    • gpt_reward_api.py Gripper manipulation and reward API.
    • grasping_utils.py Grasping utils: voxelizing, rotating, aligning grippers with Open3D.
    • motion_planning_utils.py Motion planning function for RoboGen using OMPL.
    • panda.py Panda class: Robot subclass, URDF initialization with optional params.
    • partnet_category.py PartNet dataset: 16 categories for robotic manipulation.
    • robot.py Robot control with PyBullet engine, joint and gripper management.
    • sawyer.py Sawyer robot initialization and configuration.
    • sim.py Robotics environment simulation with PyBullet, RL, and collision handling
    • table_utils.py Table manipulation utility with varying sizes and position scaling.
    • ur5.py UR5 class inherits Robot, sets up UR5 robot model.
    • utils.py Manipulation utilities for loading, rendering and simulating 3D objects.
  • objaverse_utils Objaverse utilities for mapping and object recognition.
    • embed_all_annotations.py Embeds BERT-transformed annotations for names/descriptions.
    • embed_cap3d.py Embeds object descriptions for UID mapping.
    • embed_partnet_annotations.py Embed object categories in BERT model for downstream tasks.
    • find_uid_utils.py Finds UIDs for objs in CSV using cache or search, returns similar ones.
    • utils.py Loads, checks, prints.
  • prepare.sh Prepares RoboGen environment.
  • readme.md RoboGen: AI-powered locomotion tasks execution
  • RL
    • ray_learn.py Trains RL agents, saves best models and states.
  • run.py Python script with argparse, dictionaries, folders, configs, and RL learning.