An autonomous driving agent with a Safety model and the ATtention mechanism in a multi-task framework.

CARLA version == 0.9.10.1 python >= 3.7

Anaconda is suggested for python environment management. Create the conda env with
conda env create -f environment.yml -n gym_carla

The conda env can be activated with command
conda activate gym_carla

Before you do anything, please make sure that:

1. the CARLA path in gym_carla/config/carla_config.py is set to your CARLA root path
2. the carla client is running correctly, following carla commands are suggested:
   ./CarlaUE4.sh -opengl -quality-level=Low -ResX=400 -ResY=300 -carla-rpc-port=2000
   Carla port number is alternative, however you should make sure it coordinates with the one you set in codes.

You can run the RL training with following command, and replace your_projecct_path with your own path

python your_projecct_path/SAT/rl_agents/td3_old/multi_task/developing/train_ablation.py

You can also run the script gym_carla/safety_layer/data_collector.py to collect data for the safety layer training.

Also, you can run the script gym_carla/safety_layer/train_loop.py to train the safety layer.

The safety layer model weights in our paper will be released later.

If you use this project in your work, please consider citing it with:

@article{liu2022multi,
  title={Multi-task safe reinforcement learning for navigating intersections in dense traffic},
  author={Liu, Yuqi and Gao, Yinfeng and Zhang, Qichao and Ding, Dawei and Zhao, Dongbin},
  journal={Journal of the Franklin Institute},
  year={2022},
  publisher={Elsevier}
}