On the Reliability of Spacecraft Swarms

Satellite swarms, consisting of a large number of identical, miniaturized and simple satellites, are claimed to provide an implementation for specific space missions which require high reliability. However, a consistent model of how reliability and availability on mission level is linked to costand time-effective design of the individual swarm satellites has not yet been done. We have established a method to model how applied technology and processes for designing swarm satellites under cost and time constraints impact the system-level performance of swarms. The method is applied and discussed for a future astronomy mission using a satellite swarm. Swarm satellites are severely constrained by mass, as they have to be produced in large numbers. This generally implies that they feature drastically reduced internal redundancy. This is only acceptable when all satellites are functionally identical, and can hence take over certain tasks of a malfunctioning satellite, resulting in a graceful degradation of the system performance. This swarm feature renders it significantly flexible and robust, yet it potentially affects its reliability and system throughput. Therefore, in this paper we investigated and show how the reliability of an individual satellite transfers into the overall system reliability, and hence the associated throughput of the swarm as a whole. We generated a generic model of a simple swarm satellite, and used it in conjunction with a Markov-chain based reliability analysis to assess the impact of a failure of each of the sub-systems on the functionality of the individual satellite, as well as its impact on the functionality of the swarm. Further analysis was done using Monte-Carlo simulations. The research focussed mainly on the (useful) lifetime of the system as a whole both when considering full and partial failures of elements. This was done using the assumption that swarm elements could still function in a reduced operational state when non-critical components failed. Also, the effect of recoverable malfunctions is investigated.