This repository contains code for both the F1TENTH simulation environment and the physical F1TENTH racecar, using the MIT BWSI Racecar API (racecar_core) for control and sensor interaction.
- Simulation Code: Implements wall-following using a 360 degree LiDAR in a controlled environment
- Physical Car Code: Adapts the simulation logic for a 270 degree LiDAR on the NVIDIA Jetson Xavier NX
- Clone the installer (requires Python 3.9 or lower):
git clone https://github.com/MITRacecarNeo/racecar-neo-installer.git
cd racecar-neo-installer/racecar-student/scripts
bash setup.sh
source ~/.bashrc
- Run the Simulator:
cd ~/racecar-neo-installer/racecar-student/labs
racecar sim simulation/demo.py
- Launch the GUI:
cd ~/racecar-neo-installer/RacecarNeo-Simulator
./RacecarSim.AppImage # or RacecarSim.exe on Windows
For full details and troubleshooting, refer to the MIT BWSI RACECAR Neo Installation Guide View PDF.
Install the racecar_core and utility modules:
pip install grpcio
git clone https://github.com/MITRacecarNeo/racecar-core.git
cd racecar-core
pip install -e .
In PID.py, wall-following is implemented with:
import racecar_core
import simulation.lidar_sim as ldr
rc = racecar_core.create_racecar()
lidarSim = ldr.LidarSim(rc)
# PID parameters: kp=4, kd=1
Run:
racecar sim PID.py
In car.py, use:
import racecar_core
rc = racecar_core.create_racecar(False)
# PID parameters: kp=0.02, ki=0.00001, kd=0.06
Run on the Jetson:
python3 car.py
Ensure the car is powered, and you have SSH access to the Jetson Xavier NX.
- Developed in both simulation and on Jetson Xavier NX
- LiDAR Field of View: Mapped 270 degree sensor data into a 360 degree view by padding blind spots
- Zero-degree Alignment: Determnined readings and added a steering bias offset (
+0.2). - Noise Reduction: Reduced sector width to 30 degrees and applied median filtering
- Racecar API Docs: racecar_core · MIT BWSI
- Installation Guide: Part 1/2 from MIT BWSI Autonomous RACECAR
- Alessandra Lanz
- Noah Picarelli-Kombert
- Paul Case
- Varsha Iyer Rajkumar