Cooperation Strategies for Swarms of Collaborating Robots: Analysis of Time-Stepped and Multi-Threaded Simulations

Swarms of robots have been proposed for use in many tasks, such as space exploration, search & rescue operations, and mine clearance. For a robot swarm to be successful, it needs to be self-adaptive, making its own decisions and adjusting its behaviour without relying on human intervention. This paper investigates the potential for using an autonomic system for a robot swarm engaged in a foraging task, capable of adjusting its cooperation strategy based on the ongoing performance in the task, rather than sticking with an initial strategy. The results show that while support for changing the strategy completely is limited, there remains the potential for adjusting the parameters of the given strategy to suit the ongoing situation. In addition, a comparison of two approaches to the implementation of a simulation is also presented. A time-stepped approach is compared with the multi-threaded approach used in previous work, with a view to embedding simulation within the swarm as a means of aiding the autonomic decision-making process through simulation of potential options. It is found that even when the underlying robot behaviour is identical, the time-stepped simulation is faster and more flexible, and is therefore more suitable for embedding.