.png){kind=icon}
Return
Executable Multi-Agent Path Finding with Practical Hardware Constraints
.png){kind=icon}
PRE
AI
DOI：10.1109/tase.2026.3733001.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
This paper addresses the challenges in Multi-Agent Path Finding (MAPF) for Autonomous Mobile Robots (AMRs) in logistics systems, with a particular focus on real-world constraints such as physical agent characteristics and execution uncertainties. We propose the Executable Multi-Agent Path Finding (E-MAPF) framework, which incorporates the driving structure of AMRs and considers potential collisions based on agent dimensions, trajectories, and orientations. To solve the E-MAPF problem, we develop the Executable and Continuous Conflict Based Search (E-CCBS) algorithm, which extends classical Conflict-Based Search (CBS) by integrating kinematic constraints and continuous-time execution. We further introduce the Temporal Action Plan Graph (TAPG), which executes the resulting plan through action-completion events alone and is proved to remain collision-free under arbitrary execution delays. An ablation study that isolates representation from search shows that the ability to solve E-MAPF instances stems from the orientation-dependent representation rather than from a faster search: given the same disk representation used by prior solvers, instances that E-CCBS solves become unsolvable. A demonstration on a physical testbed of eight AMRs and a mixed-fleet study with Ackermann-steering forklifts show that the representation and the event-driven execution transfer beyond the benchmark setting. This study thus bridges the gap between theoretical pathfinding solutions and their executable counterparts in event-driven fleet management systems.
Keywords:
Multi agent path finding(MAPF)
Autonomous mobile robots(AMR)
Collision avoidance
Simulation
Kinematic constraints
Journal
IF:
6.4
Papers:
5.1K
Citations:
1.6W
Organization
Cited Papers
No cited papers available