pymapf
by openplan-labs
Collection of MAPF planners and decentralized controllers for multi-agent coordination
What It Does
Implements classical and decentralized multi-agent planning and pathfinding algorithms (CBS, PIBT, LaCAM, LNS, space-time A*) plus decentralized controllers and swarm behaviors. Provides both centralized MAPF solvers and decentralized control techniques (NMPC, velocity obstacles) to simulate collision-free motion and coordinated group behaviors. Includes algorithms and reference implementations useful for reproducing planning experiments and stress-testing agent interaction patterns. Planning Pattern Event-Driven Agent Pattern
Key Benefits
Ideal For
Robotics and simulation teams needing reference implementations of MAPF algorithms to test coordination, safety, and failure modes. Agent Service Mesh Pattern
Applications
- Simulating collision-free trajectories for dozens of agents to validate coordination strategies
- Comparing centralized MAPF solvers and decentralized controllers to reveal interaction failure modes
- Stress-testing agent delegation and handoffs in physical or simulated shared spaces
- Generating reproducible scenarios for benchmarking planner performance and scalability