CRH*: A Deadlock Free Framework for Scalable Prioritised Path Planning in Multi-robot Systems

Multi-robot system is an ever growing tool which is able to be applied to a wide range of industries to improve productivity and robustness, especially when tasks are distributed in space, time and functionality. Recent works have shown the benefits of multi-robot systems in fields such as warehouse automation, entertainment and agriculture. The work presented in this paper tackles the deadlock problem in multi-robot navigation, in which robots within a common work-space, are caught in situations where they are unable to navigate to their targets, being blocked by one another. This problem can be mitigated by efficient multi-robot path planning. Our work focused around the development of a scalable rescheduling algorithm named Conflict Resolution Heuristic A* (CRH*) (https://github.com/iranaphor/crh_star) for decoupled prioritised planning. Extensive experimental evaluation of CRH* was carried out in discrete event simulations of a fleet of autonomous agricultural robots. The results from these experiments proved that the algorithm was both scalable and deadlock-free. Additionally, novel customisation options were included to test further optimisations in system performance. Continuous Assignment and Dynamic Scoring showed to reduce the make-span of the routing whilst Combinatorial Heuristics showed to reduce the impact of outliers on priority orderings.