Multi-Robot Path Planning in Narrow Warehouse Environments with Fast Feasibility Heuristics

Abstract

In this paper, we present a novel method for solving the multi-robot path planning problem in narrow warehouse environments. Robots may have to be operated in close proximity to each other in narrow lanes between pod clusters, which challenges robot-robot collision avoidance. This motivates us to consider an efficient algorithm to generate collision-free robot paths in narrow warehouse environments. The algorithm builds a two-stage scheme based on integer programming: (i) multi-robot path planning under one-way constraints; and (ii) fast feasibility heuristics. The developed algorithm has two advantages: (i) the proposed scheme successfully avoids the situation where the robots collide in narrow lanes; and (ii) by using the fast feasibility heuristics which replaces the integer programming solver's built-in feasibility heuristics, the feasibility heuristics' running time can be effectively reduced while producing better initial feasible solutions for the integer programming model solving process. Compared with the existing optimal, sub-optimal, and polynomial-complexity algorithms, our developed algorithm can leverage the advantage of integer programming to effectively reach a balance between running time and optimality in narrow warehouse environments. This point is demonstrated in the simulations.