Search and restore: a study of cooperative multi-robot systems

Swarm intelligence is the study of natural biological systems with the ability to transform simple local interactions into complex global behaviours. Swarm robotics takes these principles and applies them to multi-robot systems with the aim of achieving the same level of complex behaviour which can result in more robust, scalable and flexible robotic solutions than singular robot systems. This research concerns how cooperative multi-robot systems can be utilised to solve real world challenges and outperform existing techniques. The majority of this research is focused around an emergency ship hull repair scenario where a ship has taken damage and sea water is flowing into the hull, decreasing the stability of the ship. A bespoke team of simulated robots using novel algorithms enable the robots to perform a coordinated ship hull inspection, allowing the robots to locate the damage faster than a similarly sized uncoordinated team of robots. Following this investigation, a method is presented by which the same team of robots can use self-assembly to form a structure, using their own bodies as material, to cover and repair the hole in the ship hull, halting the ingress of sea water. The results from a collaborative nature-inspired scenario are also presented in which a swarm of simple robots are tasked with foraging within an initially unexplored bounded arena. Many of the behaviours implemented in swarm robotics are inspired by biological swarms including their goals such as optimal distribution within environments. In this scenario, there are multiple items of varying quality which can be collected from different sources in the area to be returned to a central depot. The aim of this study is to imbue the robot swarm with a behaviour that will allow them to achieve the most optimal foraging strategy similar to those observed in more complex biological systems such as ants. The author’s main contribution to this study is the implementation of an obstacle avoidance behaviour which allows the swarm of robots to behave more similarly to systems of higher complexity.