Applying RL in autonomous driving necessitates the use of robust car-driving simulators, as the RL agent cannot be trained on the road directly. To this end, several open source car-driving simulators have been developed by the research community, with each having its own pros and cons. Some of the open source car driving simulators are:
- CARLA
- http://vladlen.info/papers/carla.pdf
- Developed at Intel labs
- Suited to urban driving
- TORCS
- http://torcs.sourceforge.net/
- Racing car
- DeepTraffic
- https://selfdrivingcars.mit.edu/deeptraffic/
- Developed at MIT
- Suited to highway driving
