If you have any problems related to the accessibility of any content (or if you want to request that a specific publication be accessible), please contact (firstname.lastname@example.org). We will work to respond to each request in as timely a manner as possible.
Efficient Multi-Robot Path Planning in Discrete Spaces
AuthorLuna, Ryan J.
AdvisorBekrs, Kostas E
Computer Science and Engineering
AltmetricsView Usage Statistics
Multi-robot path planning involves moving multiple robots from their unique starting positions to their unique goals in a common graph discretization. This problem appears in a variety of applications, including planetary exploration, warehouse management, intelligent transportation networks, assembly and disassembly, autonomous robotic mining, as well as computer games. Unfortunately, computing the optimal set of paths is NP-complete, making a naive search computationally prohibitive. Typical methods to solve this problem either consider the full composite robot, which guarantees optimality and completeness but suffers from exponential complexity, or decoupled approaches which compute very fast solutions by considering each robot individually, but suffer from deadlocks. The complexity versus completeness tradeoff makes efficient coupled and decoupled solutions highly desirable, depending on the problem domain. This work presents three novel techniques for effectively solving multi-robot path planning instances. The first is related to intelligent transportation networks, and utilizes a distributed sensor network to compute and re-plan paths in a decoupled manner for hundreds of robots in real-time. The second algorithm is a centralized method that provides completeness for a sub-class of multi-robot problems by removing dependencies between robots using intermediate goals, and quickly solves the problem using a sequence of single robot paths. The third technique, also centralized, is complete for a more general class of problems, and employs sequential "push" and "swap" primitives to solve instances faster than state-of-the-art coupled and decoupled approaches.