How joonhyung-lee/mujoco-robotics-usage works
Project type and purpose
• This is an open-source robotics simulation/demo toolkit built on DeepMind’s MuJoCo engine.
• It provides hands-on Jupyter notebook demos for manipulator kinematics, trajectory planning, mobile planning, mobile-manipulator control, point-cloud processing and miscellaneous utilities.
File structure
• Top‐level “asset” directory: contains all MJCF/URDF robot and scene definitions (Husky, UR5e, Panda, realworld objects, Snapbot, etc.) plus meshes.
• “code” directory subdivided by domain:
• manipulator (inverse kinematics & trajectory planning demos)
• mobile (Husky motion planning)
• mobile_manipulator (Husky + UR5e whole-body demos)
• point-cloud (projection, voxelization demos)
• miscellaneous (viewers, segmentation demos)
• “utils” directory: parser and helper functions to load MJCF/URDF, wrap MuJoCo API, common math routines.
• README provides high-level description of algorithms, external dependencies (MuJoCo, mujoco-python-viewer, UCN, SAM).
README for additional insights
• Describes supported algorithms (IK variants, trajectory planners, global/local planners, point-cloud methods, segmentation integration).
• Lists external libraries (DeepMind MuJoCo, YAMT, MuJoCo-menagerie, UCN, SAM).
Elements
• Notebook → Demo Module → Utils → Engine → Assets (load)
• Engine → Viewer (render)
• Engine → Demo Module (state readout)
• Demo Module → External Libraries (segmentation)
• Red clouds for External Libraries