SMART: Scalable Multi-Agent Realistic Testbed

SMART is a scalable multi-agent realistic testbed for evaluating Multi-Agent Path Finding (MAPF) algorithms in settings that better reflect real robot execution. It connects MAPF planners with physics-engine-based simulation and an execution monitoring framework based on Action Dependency Graphs.

Highlights

• Published in IEEE Robotics and Automation Letters (RA-L), 11(6), 7428-7435, 2026
• ICAPS 2025 Best Demo Award
• Will present at IROS 2026
• Scales to thousands of robots while modeling kinodynamics and execution uncertainties

Links

• Online Interface
• Paper
• Publication Entry
• Code
• Video

Collaborators

Jingtian Yan, Zhifei Li, William Kang, Kevin Zheng, Yulun Zhang, Zhe Chen, Yue Zhang, Daniel Harabor, Stephen F. Smith, Jiaoyang Li