If you use this code, please cite our following paper
@inproceedings{ding2020task, title={Task-motion planning for safe and efficient urban driving}, author={Ding, Yan and Zhang, Xiaohan and Zhan, Xingyue and Zhang, Shiqi}, booktitle={2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)}, pages={2119--2125}, organization={IEEE} }
- install CARLA (0.9.10.1)
- ubuntu (18.04)
- python (3.5)
- change weather
- some weather to select: ClearNoon, ClearSunset, CloudyNoon, CloudySunset, Default, HardRainNoon, HardRainSunset, MidRainSunset, MidRainyNoon, SoftRainNoon, SoftRainSunset, WetCloudyNoon, WetCloudySunset, WetNoon, WetSunset
- change town map
- some map to select: Town01, Town02, Town03, Town04, Town05
python config.py --map town05 --weather ClearNoon --delta-seconds 0.05
- create facts to be used in task planner according to the town map
- Note(*), it seems the town05 map has been updated by adding more road id. Because some road_id did not exist before. Thus, we still use the previous "waypoints_info_sorted_previous.txt"
- compute the cost and initiate risk value of each lane
- compute an optimal task plan with input source and destination
- callback ASP code (facts.asp; problem.asp; rulesDriving.asp)
where wayPoints.npy stores location of all waypoints and numLane.npy stores the cost of each lane
- compute a motion plan with input task plan
Output: "interaction/coords_motionPlanner.txt"; "interaction/wayPoints.npy"; "interaction/numLane.npy"
- automoatically run TMPUD project
- run our ego car following task and motion planning
- play recording video stored in "/saved"