Robot Brownian Simulation

A seeded Brownian-motion simulation for multiple robots, visualized with Bevy. Each robot publishes its pose through a separate ROS 2 node, and the simulator can launch a distributed trustworthiness checker that reallocates monitoring responsibilities as robots move.

Requirements

• Rust and Cargo
• ROS 2 Jazzy
• The robosapiens-trustworthiness-checker repository for distributed monitoring
• uv for scalability benchmark sweeps

The checker is expected at ../robosapiens-trustworthiness-checker by default. Use --trustworthiness-checker-dir and --trustworthiness-checker-ros-setup if it is installed elsewhere.

ROS 2 must be sourced before compiling because r2r reads the ROS environment at build time, including for runs that use --no-ros.

Run the simulation

source /opt/ros/jazzy/setup.bash
cargo run -- --robots 20 --seed 123 --brownian-scale 35

To run the visualization without publishing poses:

source /opt/ros/jazzy/setup.bash
cargo run -- --no-ros --robots 20 --seed 123

A seed and simulation configuration produce the same trajectory. Robots use Brownian motion with inward drift near arena walls.

Run distributed monitoring

source /opt/ros/jazzy/setup.bash
cargo run -- --robots 4 --trustworthiness-checker

--tc is an alias for --trustworthiness-checker. The simulator generates the checker configuration, starts one worker per robot and one scheduler, and uses quadrant-based constraints to assign machine properties to workers. Generated configuration and logs are written under target/trustworthiness-checker/.

Each robot publishes geometry_msgs/msg/Pose2D on its own topic:

/robot_0/pose2d
/robot_1/pose2d
...

For example:

ros2 topic echo /robot_0/pose2d

Scalability benchmarks

The headless benchmark mode exercises the same simulation, ROS publishing, and checker integration while writing JSONL data:

source /opt/ros/jazzy/setup.bash
cargo run --release -- \
  --robots 10 \
  --trustworthiness-checker \
  --benchmark-duration-secs 120 \
  --benchmark-warmup-secs 20 \
  --benchmark-output-dir target/scalability-benchmarks/manual

The Python harness runs multi-size sweeps and produces Parquet summaries and plots. A short simulator-only run is:

uv run benchmark-scalability \
  --robots 2,4 \
  --seeds 123 \
  --duration 5 \
  --warmup 1 \
  --no-ros \
  --profile dev

Run cargo run -- --help and uv run benchmark-scalability --help for the full set of options.

License

This project is distributed under the INTO-CPS Association Public License (ICAPL) version 1.0. The selected usage mode is GPL; see LICENSE.md and ICA-USAGE-MODE.txt.