Reproduce Experiments

Prepare

1. Ensure you have a Gurobi license;

2. Install all dependencies by python package manager (e.g., uv): uv sync

Generate scenario files

• python gen.py special

  • Generate scenarios for experiment 1, 2 and 3
  • Each scenario is a representative scenario and has only one instance
  • All generated scenarios are store in folder special-mapf-scen

• python gen.py large

  • Generate scenarios for experiment 4 and 5
  • Each scenario has 10 instances
  • All generated scenarios are store in folder large-mapf-scen

Run experiment

Run MICP, MILP, MINLP

• Experiment 1:
  • python expr.py -e special -dt 0.2 -t 10 -a micp
  • python expr.py -e special -dt 0.2 -t 10 -a milp
  • python expr.py -e special -dt 0.2 -t 10 -a minlp
• Experiment 2:
  • python expr.py -e large -dt 0.2 -t 10 -a micp
  • python expr.py -e large -dt 0.2 -t 10 -a milp
  • python expr.py -e large -dt 0.2 -t 10 -a minlp

Result of instances are store in the corresponding data directory, e.g. large-mapf-scen/n4-10x10-swap/0-micp-UB.json.

Reference

@inproceedings{micp_mamp,
  author = {Shizhe Zhao, Yongce Liu, Howie Choset and Zhongqiang Ren},
  booktitle = {2025 IEEE/RSJ International Conference on Intelligent Robots and Systems},
  title = {Mixed Integer Conic Programming for Multi-Agent Motion Planning in Continuous Space},
  year = {2025},
  code = {https://github.com/rap-lab-org/public_MICP_MAMP}
}