A Darwinian Swarm Robotics Strategy Applied to Underwater Exploration

Abstract

This work focuses on the development of an autonomous multi-robot strategy to explore unknown underwater environments by collecting data about water properties and the existence of obstacles. Unknown underwater spaces are hostile environments whose exploration is often a complex, high-risk undertaking. The use of human divers or manned vehicles for these scenarios involves significant risk and enormous overheads. The systems currently employed for such tasks usually rely on remotely operated vehicles (ROVs), which are controlled by a human operator. The problems associated with this approach include the considerable costs of hiring a highly trained operator, the required presence of a manned vehicle in close proximity to the ROV, and the lag in communication often experienced between the operator and the ROV. This work proposes the use of autonomous robots, as opposed to human divers, which would enable costs to be substantially reduced. Likewise, a distributed swarm approach would allow the environment to be explored more rapidly and more efficiently than when using a single robot. The swarm strategy described in this work is based on Robotic Darwinian Particle Swarm Optimization (RDPSO), which was initially designed for planar robotic ground applications. This is the first study to generalize the RPSO algorithm for 3D applications, focusing on underwater robotics with the aim of providing a higher exploration speed and improved robustness to individual failures when compared to traditional single ROV approaches.