Decentralized Deadlock Prevention for Self-Organizing Industrial Mobile Robot Fleets

Industrial systems more and more integrate fleets of autonomous mobile robots to perform transport tasks on the shop floor. Nowadays, central fleet management systems control all functionalities like task allocation, path planning and motion planning, which includes collision and deadlock prevention. But with increasing demand for flexibility and the ability to scale, today’s centralized systems reach technical limits. Thus, a trend towards decentralized systems can be observed. Functionality that used to be located in a central unit is delegated to the executing robots, such that every entity in the system gets more intelligent and autonomous.This work targets systems that are in need of a decentralized algorithm for preventing collisions and deadlocks at the start and end points of planned routes. To that end, we presuppose robots that are already able to navigate requested paths and avoid collisions during movement. We propose a new solution to this problem and compare it to an existing decentralized algorithm in a simulated environment. For that we provide requirements, qualitative and quantitative metrics and an evaluation of both algorithms for real-world industrial implementation. Recommendations are given on which algorithm to use given the target objectives of the system.