Guidelines for improving the robustness of swarm robotic systems through adjustment of network topology

Abstract

Swarm Robotics (SR) is expected to have a significant impact on society over the next decade. One of the contributing factors is that swarms are robust. However, robustness has not gained sufficient attention in the context of robotic swarms. This study focuses on the swarm network to generate insights as to how network topologies can be controlled to improve the robustness of SR systems. More specifically, how removing key robots alters the network topology, thereby changing the performance of the swarm. Analyzing these changes provides possible guidelines to improve swarm robustness towards targeted interventions. The most important findings suggest that robustness can be increased by making the network topology: (1) provincial and decentralized in the middle phase of the swarming procedure in unimodal domains, (2) provincial and centralized during the same phase in multi-objective domains.