協調配送問題のための分散ナビゲーション

抄録

Multi-agent Pickup and Delivery (MAPD) is a problem to compute non-colliding paths for multiple mobile agents to complete a stream of delivery tasks. MAPD models the actual delivery systems like automated warehouse, on-demand bus, and so on. Unlike existing approaches which solve the problem by a central computer or a set of autonomous agents, we solve MAPD problem based on a concept of active environment: the infrastructure assisted navigation system with spatially distributed computing units. By this concept, we aim to make a scalable system without single point of failure, while taking over the advantage of centralized systems. In this paper, we develop a distributed algorithm extending TSWAP, an online method for MAPD, and give a proof that the algorithm eventually completes all tasks. We also evaluate the proposed algorithm by a computer simulation.