Performance Evaluation of a Quorum Sensing based Scheme in Multi-Agent Task Development

Abstract

Robotics is positioned today as a fundamental tool in industrial and commercial development, where machines interact directly with humans. There is a vast variety of tasks that require autonomous, robust, and high-performance systems. Among these tasks can benefit from the autonomous integration of multiple elements, known as multi-agent systems. These schemes have interesting advantages over the single robot solution centered on the high degree of robustness achieved and the lower cost. The control of these multi-agent systems turns out to be of great complexity and is an active field of robotics research. The motion coordination schemes are complex and require a certain level of processing and communication. In this paper, a decentralized coordination scheme for low-cost robot groups based on local interaction is evaluated. The algorithm uses bacterial Quorum Sensing (QS) as a behavioral model, a scheme under which certain actions are triggered by the agents conditioned to the population density in the region they cover. The algorithm is tested in navigation tasks for different conditions of the design parameters. Among the parameters evaluated are environment dependence, system size, and QS threshold. The development times of the tasks were statistically analyzed, and a strong dependence of the environment on the total time required was found (a well-structured and small environment concerning the system improves the performance considerably), as well as the design of the robot in terms of QS threshold and sensors.