Growing Dependability using a Multi-Agent Approach to Fault Tolerance
Peter Mendham · 54th International Astronautical Congress of the International Astronautical Federation, the International Academy of Astronautics, and the International Institute of Space Law · 2003
The use of Artificial Intelligence (AI) in creating autonomous, fault tolerant systems is becoming increasingly important for complex space software. Recent applications of a single intelligent agent in deep space probe management have highlighted the potential of AI in space applications. Whilst highly suited to small spacecraft, this type of approach can represent a single point of failure and is less than ideal for highly complex systems with distributed centres of computing and without the potential for standby operating states. We present an argument in favour of software built from multiple agents as a useful representation of a complex, fallible system. We propose a novel approach to the development of highly dependable space systems software using multiple agents to react cooperatively and autonomously to fault conditions. To further explain our approach, we discuss the system in the context of the Skylon spaceplane, an autonomous SSTO launch vehicle, highlighting the need for fault tolerance and dependability within hard real-time constraints. Finally, we discuss the challenges in applying multiagent systems more widely in the software of complex space systems.