Chris Urmson is often credited for the phrase “simulations are doomed to succeed” to outline the importance of simulations in self-driving cars and autonomous systems in general. There are many simulators out there however only a few go beyond very basic environments that are made for toy projects/course work. There are a few (open source) simulators out there that go beyond that. Most of them are based on Unity 3D or Unreal engines. I haven’t seen anything based on Blender even though I would like to see that due to less licensing issues.
Drones/Air transport
- Microsoft AirSim
- Udacity Flying Car Simulator
- Perhaps FlightGear might be possible as well since it is open-source and simulators build on top of it are FAA certified (or at least some parts of them).
Self-driving cars
- CaRL
- CARLA
- Have a look at my first look on CARLA
- Udacity Self-driving Car Simulator
- TORCS via pyTORCS
- Rigs of Rods It is a game but the physics are aimed to be very realistic so it could be utilized
- Apollo seems to have some very limited simulation capabilities but I’m unsure if we can count that as a simulator.
- Autoware
- LGSVL Simulator by LG based on Apollo and Autoware
Underwater robotics
- UUV Simulator (Unmanned Underwater Vehicle Simulator)
- built on-top of ROS
- UWSim (UnderWater SIMulator)
Miscellaneous
- FLOW
- Google Research Football
- Gymnasium
- fork that continues OpenAI Gym which is deprecated
- SUMO (Simulation of Urban MObility)