In this repository tensorflow implementation of Deep Q-Learning is used for self-driving vehicle in CARLA environment. The algorithm is implemented in a quite simple environment with few surrounding vehicles. An example of the result can be seen below. Note that the agent requires to be trained longer than the figure provided with more obstacles on the road.
Clone the repository git clone https://github.com/shayantaherian/Reinforcement-Learning-CARLA/.git
Install the requirement requirement.txt
Download Carla you can just download the compiled version from here. Note that it is reuiqred to download the stable version of the simulator
First run the carla server CarlaUE4.sh from the save directory
Then run python Main.py. To test the results run python Test.py
Note that to add more vehicle into simulation run py -3.7 spawn_npc.py -n # which # is the number of surrounding vehicle