Multi-objective collective search and movement-based metrics in swarm robotics

Abstract

Particle Swarm Optimization is a well researched meta-heuristic for single- and multi-objective problems. It is based on the movement of particles, which enables its application to collective search in robotic applications. However, in the robotic context some assumptions regarding performance measurement of PSO-algorithms do not apply. Concerning energy and time, while reading a sensor is usually cheap, the movement is costly in robotic applications. Traditional methods do not take into account any cost metrics associated with the actual movement of the particles. Therefore, new metrics are required to understand how well a PSO-algorithm can perform as a collective search mechanism. This article proposes two metrics, the Normalized Movement Energy Cost and Normalized Movement Time Cost that enable researchers to analyze an algorithm's performance not just regarding the obtained solution quality, but also with respect to movement time and energy costs.