Documentation

Simulation

Two ways to run the full stack with no robot on the desk: the MuJoCo example for the whole plan-and-execute loop, and URSim for the real UR bridge.

#MuJoCo ReBeL example

The example in examples/mujoco_rebel drives a simulated igus ReBeL to a joint goal. Its graph is only library components (goal → RRT* planner → timing → follower → controller, diagram in Architecture); the single application node is an observer that writes result.json and exits non-zero on timeout or controller fault. The bridge, repose_mujoco_sim::CuMujocoSimBridge (umbrella feature mujoco-sim), talks line-delimited JSON over TCP to a Python sidecar that steps MuJoCo physics; only command positions cross the wire, state comes back as q/qd in radians.

python -m pip install mujoco numpy
python sidecar.py --urdf /path/to/igus_rebel.urdf   # terminal 1
cargo run --manifest-path Cargo.toml                 # terminal 2, from examples/mujoco_rebel

• Requirements: Rust 1.95+, Python 3 with mujoco and numpy, and an expanded ReBeL URDF whose movable joints are named joint1…joint6 in chain order. The URDF supplies MuJoCo's inertial model — verify it describes the same robot as the igus_rebel preset it is checked against.
• Configuration: goal, planner budgets/seed and motion limits are all in its copperconfig.ron (bridge default 127.0.0.1:55091 there, 8765 in the crate default; change both sides together). The run writes a replayable sim_demo.copper log; planning failures appear as a published CuPlanStatus.
• Limits: a demonstration, not a hardware interface — collision checks are discrete, and nothing monitors collisions during execution.

#URSim

For the UR bridge against Universal Robots' own simulator, follow Run on URSim: pinned URSim 5.25.1 in Docker, PolyScope setup with Remote Control, then the same robot-app template with robot: ur5e. Everything verified there — joint-state feedback and streamed multi-step programs — is verified on the simulator only.

#Which one?