Ambiguous Fault Diagnosis in Robotic Swarms
Abstract: Robotic swarms can suffer from seemingly endless technical issues when applied in real-world applications. This is because robots become less reliable as parts of their system fail over repeated use. This issue is exacerbated in swarms as there are multiple robots all failing at different times. A desirable situation is a swarm of robots that can identify and handle any faults as they arise to prevent any negative effects on performance. This means that agents will need to diagnose robots based on minor differences and understand the unique characteristics of robots experiencing detrimental faults. A swarm capable of accurately diagnosing faulty agents could be capable of self-maintenance, which will mitigate these issues significantly, making the swarm more robust and capable. My work aims to develop a fault detection and diagnosis system that individuals in a swarm can use to diagnose ambiguous faults in other agents.
Advisor: Professor Carlo Pinciroli
Committee: Professor Kevin Leahy and Professor Nitin Sanket