Scaling up Multi-Agent Systems through Organizational Structuring
Victor Lesser · Web Intelligence · 2004
Over the last four years, I have been working on the design of a distributed sensor network for vehicle tracking using adaptive radar sensors. This effort has re-ignited my long-term interest in the use of organizational structuring for handling scaling issues in multi-agent systems. In this lecture, I will first examine how organizational structuring was used in our design to reduce the computational and communication load involved in coordinating agent activities, and I will discuss experiments that indicate the complex set of issues that need to be considered in evaluating the effectiveness of different organizational variants. I will also briefly illustrate how it is possible to model the performance of these variants analytically, and predict their performance. I will then present bottom-up techniques, based on negotiation, for organizational instantiation and adaptation in a dynamic environment. Finally, I will discuss some recent work on a knowledge-based top-down approach for organizational design.