A neural network approach to monitoring robot malfunction in multirobot formation control tasks

This paper presents a design of a high-level monitoring system for formation controls of swarms of mobile robots. With necessary information of each robot provided in the centralized high-level planner, the proposed monitoring system can evaluate the robots' working performance and check which robot malfunctions. The malfunction detector, developed with back-propagation neural networks technology, helps make decision whether the robots fails to meet the task requirement and should be abandoned from the team. The neural network is trained using both positive and negative examples. Case studies are performed to demonstrate the effectiveness of the proposed approach.