COORDINATION IN MULTI-AGENT ROBOTIC SYSTEMS

Abstract

During last few decades, major research efforts were focused on improving the performance of an individual mobile robot by using advanced sensors, actuators, and intelligent control algorithms. This was mainly driven by the need to perform increasingly complex tasks required by real world applications. As a result, individual mobile robot has become very sophisticated. The current trend in robotic research, both from engineering and behavioral viewpoints, has been to move away from the design and deployment of few, rather complex, usually expensive, application-specific robots. In fact, in the last decade the interest has shifted towards the design and use of a large number of “generic” robots which are very simple, with very limited capabilities and, thus, relatively inexpensive, but capable, together, of performing rather complex tasks. In a system consisting of a set of totally distributed agents the goal is generally to exploit the multiplicity of the elements in the system so that the execution of a certain predetermined task occurs in a coordinated and distributed way. The advantages of this approach are clear and many, including: reduced costs (due to simpler engineering and construction costs, faster computation, development and deployment time, etc); ease of system expandability (just add a few more robots) which in turns allows for incremental and on-demand deployment (use only as few robots as you need and when you need them); simple and affordable fault-tolerance capabilities (replace just the faulty robots); re-usability of the robots in different applications (reprogram the system to perform a different task). Moreover, tasks that could not be performed at all by a single agent become manageable when many simple units are used instead. I intend to design and implement a distributed multi-robot system using some multirobot simulation tool. Coordinating robots in a dynamic environment is a difficult task. They must be able to carry out their contributions to the overall goal of the system efficiently and effectively while not impeding each other. The focus of multi-robot coordination should therefore be twofold: each robot should consider the objectives of the

iii team while maintaining its own functional integrity. As such, the team plan should exist at a level where it provides strategies for each robot to contribute to the team’s success. Each robot must consider the team strategy and execute it as best it can without compromising its ability to maintain functional operation. There are two ways of making the robots coordinate. First is to use explicit communication between the robots so that they may easily coordinate. But in actual scenarios this explicit communication is usually not possible or feasible (e.g. in a hostile environment). The other choice is to make the robots coordinate by observing their environment or the behaviors of the other robots. This latter approach is a new concept in recent research. I wish to study and implement this approach during the course of my thesis work.