yding25/TMPUD

TMPUD

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} }

Environment

• install CARLA (0.9.10.1)
• ubuntu (18.04)
• python (3.5)

Config environment

config.py

function

• 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

prerequisite: run ./CarlaUE4.sh

input: None

output: None

example:

python config.py --map town05 --weather ClearNoon --delta-seconds 0.05

Task Planner

facts_from_townmap.py

function

• create facts to be used in task planner according to the town map

prerequisite: run ./CarlaUE4.sh

input: None

output: "task-level/all_waypoints_list_sorted.txt"

example: python facts_from_townmap.py

• 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"

get_cost_risk_of_lane.py

function

• compute the cost and initiate risk value of each lane

prerequisite: None

Input: "task-level/all_waypoints_list_sorted.txt"

Output: "interaction/lanes_cost.txt"; "interaction/lanes_risk.txt"

example: python get_cost_risk_of_lane.py

get_optimal_task_plan.py

function

• compute an optimal task plan with input source and destination
• callback ASP code (facts.asp; problem.asp; rulesDriving.asp)

Input: source and destination; "interaction/lanes_risk.txt"; "interaction/lanes_cost.txt"

Output: "interaction/task_plan.npy";

where wayPoints.npy stores location of all waypoints and numLane.npy stores the cost of each lane

example: python get_optimal_task_plan.py 37 47

ground_task_plan.py

function

• compute a motion plan with input task plan

Input: "interaction/task_plan.npy"; "task-level/all_waypoints_list_sorted.txt"

Output: "interaction/coords_motionPlanner.txt"; "interaction/wayPoints.npy"; "interaction/numLane.npy"

example: python ground_task_plan.py

TMPUD

autorun.py

function

• automoatically run TMPUD project

main.py

function

• run our ego car following task and motion planning

replay.py

function

• play recording video stored in "/saved"