Swarm shape manipulation through connection control

The control of a large swarm of distributed agents is a well known challenge within the study of unmanned autonomous systems. However, it also presents many new opportunities. The advantages of operating a swarm through
distributed means has been assessed in the literature for efficiency from both operational and economical aspects; practically as the number of agents increases, distributed control is favoured over centralised control, as it can reduce agent computational costs and increase robustness on the swarm. Distributed architectures, however, can present the drawback of requiring knowledge of the whole swarm state, therefore limiting the scalability of the swarm.

In this paper a strategy is presented to address the challenges of distributed architectures, changing the way in which the swarm shape is controlled and providing a step towards verifiable swarm behaviour, achieving new configurations, while saving communication and computation resources. Instead
of applying change at agent level (e.g. modify its guidance law), the sensing of the agents is addressed to a portion of agents, differentially driving their behaviour. This strategy is applied for swarms controlled by artificial potential functions which would ordinarily require global knowledge and all-to-all
interactions. Limiting the agents’ knowledge is proposed for the first time in this work as a methodology rather than obstacle to obtain desired swarm behaviour.