A Feedback Scheme to Reorder a Multi-Agent Execution Schedule by Persistently Optimizing a Switchable Action Dependency Graph

In this paper we consider multiple Automated Guided Vehicles (AGVs) navigating a common workspace to fulfill various intralogistics tasks, typically formulated as the Multi-Agent Path Finding (MAPF) problem. To keep plan execution deadlock-free, one approach is to construct an Action Dependency Graph (ADG) which encodes the ordering of AGVs as they proceed along their routes. Using this method, delayed AGVs occasionally require others to wait for them at intersections, thereby affecting the plan execution efficiency. If the workspace is shared by dynamic obstacles such as humans or third party robots, AGVs can experience large delays. A common mitigation approach is to re-solve the MAPF using the current, delayed AGV positions. However, solving the MAPF is time-consuming, making this approach inefficient, especially for large AGV teams. In this work, we present an online method to repeatedly modify a given acyclic ADG to minimize route completion times of each AGV. Our approach persistently maintains an acyclic ADG, necessary for deadlock-free plan execution. We evaluate the approach by considering simulations with random disturbances on the execution and show faster route completion times compared to the baseline ADG-based execution management approach.