Anticipating Degradation: A Predictive Approach to Fault Tolerance in Robot Swarms

摘要

An active approach to fault tolerance is essential for robot swarms to achieve long-term autonomy. Previous efforts have focused on responding to spontaneous electro-mechanical faults and failures. However, many faults occur gradually over time. This work argues that the principles of predictive maintenance, in which potential faults are resolved before they hinder the operation of the swarm, offer a promising means of achieving long-term fault tolerance. This is a novel approach to swarm fault tolerance, which is shown to give a comparable or improved performance when tested against a reactive approach in almost all cases tested.