Towards a semantics of desires (abstract)

Research in Distributed Artificial Intelligence (DAI) often considers the problem of how best to utilize multiple automated agents to accomplish a given task. A canonical problem, the Pursuit Problem (1), has been suggested as a useful tool for evaluating alternative approaches to the distribution of knowledge and control among intelligent, cooperative problem-solvers. Work on the Pursuit Problem has been carried out by several researchers; for example, Stephens and Merx (16, 171 compare alternative approaches to solving the problem.In [5], a game theoretic model was proposed for DAI in which each agent works for his own selfish goals. In this paper, we suggest a method for solving the Pursuit Problem using game theoretic techniques, by incorporating the global goal of a group of agents into their local interests. Although applied to the Pursuit Problem, the technique has wider applicability throughout DAI. We present the results of a simulation of the game theory model for the Pursuit Problem, and compare results to those of Stephens and Merx's models.