Multiagent Systems for Environmental Perception

Multiagent systems are distributed systems in which a number of individual autonomous heterogeneous agents interact and operate in an environment. Agents can be physical (computers or robots) or logical (software) entities. One of the early classical applications of the multiagent systems was in distributed sensing and monitoring, as appears from the work by Victor Lesser and colleagues (1981) on the Distributed Vehicle Monitoring Testbed (DVMT). Starting from this pioneering work, a number of techniques for multiagent-based environmental perception have been developed. In this paper, we discuss the state of the art and the main techniques used in multiagent systems in relation to their possible application to environmental perception. Moreover, we set off some issues that must be addressed for developing multiagent perceptive systems. This paper aims to pose problems and to present general ideas rather than giving definite answers and assessed results. In this sense, we do not intend to provide a survey of the current state of the art in the field of distributed environmental perception and, as a consequence, the works cited here are those that (to the best of our knowledge) we deem most suitable to support our argumentation. Moreover, since the field of distributed environmental perception is definitely multifaceted and since we have not competences in all the involved disciplines, we concentrate on the architectural software characteristics of the multiagent perceptive systems. The need for such system wide architectures is expressed in a very clear way by Estrin et al. (2002). Finally, we remark that the work presented here differs from that reported by Amigoni et al. (2002) because the latter provides a general and unitary framework in which an historical overview of the advent of multiagent systems in measurement science can be settled, whereas this paper is more oriented on some specific problems that arise when multiagent systems for environmental perception are implemented. This paper is organized as follows. In the next section, we shortly present the field distributed artificial intelligence and the DVMT application. Section 3 surveys the concepts and the methods of multiagent systems. In Section 4, the application of multiagent systems to environmental perception is illustrated. In Section 5, we made this discussion more practical by describing a case study of application of multiagent systems to environmental perception. Finally, Section 6 concludes the paper.