Energy-Based Particle Swarm Optimization: Collective Energy Homeostasis in Social Autonomous Robots

Abstract

Social robots, both when acting individually and in groups, need to manage their own energy use effectively. An uneven energy distribution in a swarm of robots performing a mission may prevent completion of the group's mission or reduce its tolerance for adverse events that occur during mission execution. The goal of effectively managing energy use in a swarm is known as collective energy homeostasis. While previous works have mainly focused on achieving this goal by direct energy exchange methods (e.g., using battery charging mechanisms), this paper presents a novel bio-inspired approach for maintaining collective energy homeostasis in social robot swarms. The approach extends Particle Swarm Optimization (PSO) techniques for task selection and motion planning for individual robots by making them sensitive to the robot's energy state. The overall effect of this Energy-based PSO (EPSO) algorithm is to shift energy-intensive tasks towards robots in the swarm that have higher energy levels, i.e. energy load-leveling, which improves energy self-sufficiency and its homogeneity across the swarm. This can be considered a form of indirect energy exchange by task shifting. Experimental results show that the EPSO algorithm enables social robot swarms to maintain collective energy homeostasis more effectively than previous approaches, reducing the variance of energy between individuals by 49%, and extending the number of missions that a swarm can achieve, given a fixed initial energy budget.