This program controls the actuators of a formula student driverless racing car with a machine learning technique named learning by imitation.
Firstly, you have to drive the car, preferably with a racing wheel, in order to take data of your driving. With enought collected data, next step is to train a machine learning algorithm. The testing takes place in the Webots simulator.
For further information, please check the documentation of the degree final project (awaiting publication).
The set-up used to run the executions will be shown below.
- Windows 10
- Webots R2021b
- Weka
- Python3
- Replicate the set-up
- Clone this repository
- Open
Webots >> worlds >> city_mad.wbt
- Open with the
Text Editor
Webots >> controllers >> autonomous_vehicle >> autonomous_vehicle.c
- Press
Build the current project
- Press
Run the simulation in real-time
- Different circuits and a creation tool
- Oval
- Skidpad
- FSG
- Racing Wheel PC implementation to take data
- Basic example running and the data
The car is able to complete a lap in the oval circuit with the MLP algorithm, you can easily improve the results by using different machine learning algorithms.
mlp_sim_2.mp4
This project was developed by Alejandro Parrado Arribas as part of his work at MAD Formula Team.