Evolving Cooperation Strategies

We propose an approach to developing cooperation strategies for multiagent problem solving situations which is different from existing techniques in two ways: strategies are incrementally constructed by repeatedly solving problems in the domain, i.e., on-line; and, we utilize an automated method of strategy formulation and modification, that relies little on domain details and human expertise, and more on performance on randomly generated problems in the domain. The genetic programming (GP) (Koza 1992) paradigm used to develop, through repeated problem solving, increasingly efficient strategies. Populations of structures are represented as Lisp symbolic expressions (Sexpressions). These are manipulated to evolve better structures by propagating and combining parts of structures that perform well compared to others in the population. To use this approach, we have to develop an encoding of strategies as S-expressions and choose an evaluation criterion for a strategy corresponding to an arbitrary S-expression. Strategies are evaluated by allowing the agents to execute them in the application domain and by measuring their efficiency and effectiveness by a set of criteria relevant to the domain.