/Learning-by-Imitation-FS

Development of an artificial vision system for autonomous driving in Formula Student. Public version

Primary LanguagePythonMIT LicenseMIT

banner

Learning by Imitation FS

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).

Set-Up

The set-up used to run the executions will be shown below.

How to run

  1. Replicate the set-up
  2. Clone this repository
  3. Open Webots >> worlds >> city_mad.wbt
  4. Open with the Text Editor Webots >> controllers >> autonomous_vehicle >> autonomous_vehicle.c
  5. Press Build the current project
  6. Press Run the simulation in real-time

Features

  • Different circuits and a creation tool
    • Oval
    • Skidpad
    • FSG
  • Racing Wheel PC implementation to take data
  • Basic example running and the data

Example

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

Credits

This project was developed by Alejandro Parrado Arribas as part of his work at MAD Formula Team.

Note: This public version might have some features removed MADFT_P_Main_crop