Setting-Up-CARLA-Reinforcement-Learning
Setting-Up-CARLA-Reinforcement-Learning copied to clipboard
Run training faster
Hi,
Is it possible to run the simulator faster for training the RL agent? It is too slow in real time
It is true that simulation is costly and takes time. These are some factors that affect the simulation:
- Resolution of RGB camera
- Number of cameras
- Additional outputs like depth maps and segmented images (if you're using)
- Reduce rendering resolution or Disable rendering
In addition, you can also try playing with the FPS rate in the config.py
, but it won't help if the bottleneck is your GPU and the actual simulation/graphics. Also, you may try using Carla latest versions, but I haven't added support for it / tested it.