Data fusion using large multi-agent networks: an analysis of network structure and performance

Concerns the multi-agent paradigm for structuring the data fusion in large networks of information sources. The general goal is the maximisation of the quality of the result at minimum cost. To this end agents compete for scarce resources or they work in parallel. Both the general granularity of the agent society and the competence assigned to each individual agent determine the information flow. The multitude of parameters involved makes it difficult to optimally adapt the network structure to a given class of sensing tasks. We outline possible network structures and present an approach for analysing a number of important parameters characterising the network. The abstraction enables comparison of different structures. The methods for the analysis may be readily refined to evaluate a specific problem. A model of lateral coordination control in proposed as a result. It is based on the notion of negotiated cooperation between pairs of autonomous sensor agents. The cooperation phase is preceded by a bidding scheme to establish logical communication links. The cooperation is modelled on human social behaviour