Bartolo1024/TORCS-behavioral-cloning

TORCS supervised learning agent using data from keyboard or g27 controller

Torcs Human Policy

Environment collect players input and sensors data from race. HumanAgent allows collect input data from keyboard or g27 logitech controller. Collected date can be used in supervised training. Trained agents are based on MLP and LSTM eural networks

Requirements

Project is develop using Ubuntu 16.04 LTS. Code is based on:

• Python 3
• Torcs 1.3.7
• gym-torcs (https://github.com/ugo-nama-kun/gym_torcs)
• Tensorflow
• numpy
• h5py
• g27 (https://github.com/felixhummel/g27)
• keyboard