Race stack for F1Tenth Autonomous Racing

Updates:

• We secured 5th position in the Roboracer competition at ICRA 2025.
• One of our winning races video is here: https://youtu.be/EZ4xm3I_c7I?si=HrvqoY1bP3b0-n52

Building:

1. Clone this repo

2. This repo uses some custom message types for communication, defined here: https://github.com/vaithak/f1tenth_icra_race_msgs

   git clone git@github.com:vaithak/f1tenth_icra_race_msgs.git
   colcon build --packages-select=f1tenth_icra_race_msgs
   

3. Update the config/r3.params.yaml file by changing the location of the waypoints file included in the repo - race3-manual-fitted.csv

4. Build the package

   colcon build
   source install/local_setup.bash
   

Running

• Launch the object and tracking subsystems ros2 launch f1tenth_icra_race r3.oo.launch.py
• Launch the spliner and state machine ros2 launch f1tenth_icra_race r3.ss.launch.py
• Launch the controller ros2 launch f1tenth_icra_race r3.c.launch.py

Subsystems and workings

• Opponent Detection and tracking:
  • It uses a clustering method with an adaptive breakpoint algorithm to create bounding box
    rectangles for obstacles.
  • Uses a simplified Extended Kalman Filter for Velocity tracking (assumes a single obstacle).
  • Implemented in C++ for realtime performance of 50Hz.
  • Reference: https://github.com/kostaskonkk/datmo
• Spline-based real-time planner and state machine:
  • Real-time updating of local waypoints if an obstacle is detected in front of the car.
  • Decides which side to overtake from and fits a cubic spline to update the waypoints.
  • State machine to decide when to trail vs when to overtake
  • Reference from ForzaETH: https://arxiv.org/abs/2403.11784
• Pure Pursuit-based controller:
  • Pure Pursuit controller with sector-wise adaptive lookahead and velocity mapping.

Working demo on sim

Screencast.from.05-04-2025.07.11.06.AM.webm