Autonomous Racing System
ROS2-based multi-car autonomous racing with LIDAR-driven control, Pure Pursuit steering, and wall-following algorithms in a Dockerized F1Tenth simulation environment.
Multiple control algorithms (Pure Pursuit, Wall Following, Reactive)
Full Docker containerization with cross-platform support
Multi-car simulation with real-time collision avoidance
LIDAR-based perception and trajectory planning
A comprehensive autonomous racing system built during the Autonomous Racing Cars summer project at TU Wien. The project implements multiple control algorithms for F1Tenth scaled racing cars within a ROS2 ecosystem.
The system uses LIDAR sensor data to perceive the environment and implements several control strategies: Pure Pursuit steering for trajectory tracking, reactive wall-following for unknown tracks, and a simple raceline generator for optimized paths on known circuits. A safety node monitors for imminent collisions and can override the control commands to prevent crashes.
The entire development environment is containerized with Docker, supporting multiple platforms (Linux, macOS, Windows/WSL) with both Docker Compose and VS Code devcontainer setups.
The simulator runs the F1Tenth Gym environment with RViz2 visualization, accessible through VNC for remote development.
The project also supports multi-car simulation scenarios, enabling head-to-head racing between different control algorithms — a challenging real-time systems problem requiring careful management of sensor data, control loops, and inter-vehicle awareness.