RoboNav

A ROS 2 robotics project combining occupancy-grid mapping, reactive navigation and global path planning in CoppeliaSim.

Originally developed for MSc Autonomous Robots coursework and later cleaned to isolate the planning core, improve configuration and add reproducible offline tests.

System

CoppeliaSim
   ↓ pose + range data
ROS 2 interface
   ↓
log-odds occupancy mapper
   ↓
OccupancyGrid
   ├── Bug2 reactive navigation
   ├── Theta-style grid planning
   ├── Dijkstra
   └── RRT

The mapper uses Bresenham ray tracing to update free and occupied cells in log-odds form. Robot pose is supplied by the simulator, so this is mapping with known pose, not full SLAM.

What is implemented

• ROS 2 interface to CoppeliaSim.
• Log-odds occupancy mapping.
• Bresenham free-space/ray updates.
• ROS OccupancyGrid publication.
• Bug2 reactive navigation and goal management.
• Theta-style any-angle grid planning.
• Dijkstra and RRT planners.
• Docker-based ROS 2 Humble environment.
• Offline planner extraction for tests without a simulator.

Offline demo

python -m venv .venv
source .venv/bin/activate
pip install -r requirements-offline.txt
python -m unittest discover -s tests -v
python scripts/planning_demo.py

The demo runs the planning core on a synthetic grid and reports path length and queue pops.

Simulator workflow

docker compose up --build

The original course scene is not redistributed. CoppeliaSim must expose the ZMQ remote API and the expected robot/sensor objects must exist in the loaded scene.

Example

Historical figures are qualitative evidence only; they are not navigation-success benchmarks.

Repository structure

src/                 ROS 2 package and navigation nodes
scripts/             offline planning demo
maps/                small example maps
tests/               planner tests
evidence/            historical outputs and source manifest
PROVENANCE.md        source/cleanup notes
VALIDATION.md        executed checks and limits

Engineering improvements

• extracted the planning core for direct tests;
• removed unused environment files and caches;
• parameterised the simulator address;
• corrected invalid-input handling;
• removed unused optional dependencies.

Limitations

• Full CoppeliaSim integration has not been rerun in the cleaned environment.
• Global planning assumes a point robot and does not perform footprint inflation.
• A valid grid route does not itself guarantee collision-free physical control.
• Mapping relies on externally supplied pose and is not a complete SLAM system.
• Historical figures are not quantitative benchmarks.

See VALIDATION.md and PROVENANCE.md.