Deriving Multi-Agent Coordination through Filtering Strategies

We examine an approach to multi-agent coordination that builds on earlier work on enabling single agents to control their reasoning in dynamic environments. Specifically, we study a generalization of the filtering strategy. Where single-agent filtering means tending to bypass options that are incompatible with an agent's own goals, multi-agent filtering means tending to bypass options that are incompatible with other agents' known or presumed goals. We examine several versions of multi-agent filtering , which range from purely implicit to minimally explicit, and discuss the trade-offs among these. We also describe a series of experiments that demonstrate initial results about the feasibility of using multi-agent filtering to achieve coordination without explicit negotiation. 1 Introduction Distributed Artificial Intelligence (DAI) is concerned with effective interactions, and with the mechanisms by which these interactions can be achieved. Broadly speaking , two main approaches have been proposed in the literature. The first involves explicit coordination; agents are designed to reason about their potential interactions, and negotiate with one another as needed. A difficulty with explicit coordination and negotiation is that it can be extremely time-consuming, and in dynamic environments, agents may not be able to afford the time required. The second approach involves implicit coordination; agents are designed to follow "local" rules of behavior that lead to their acting in apparently coordinated ways; see, for example, [Shoham and Ten-nenholtz, 1992; Goldman and Rosenschein, 1994]. This approach is motivated in part by a belief that one can design simple rules that are easy for an agent to follow, yet result in coordination. In this paper, we take the second approach, examining an implicit coordination strategy. The strategy we study, multi-agent filtering, is an extension of a single-agent strategy for controlling reasoning in dynamic environments. The notion of single-agent filtering derives from the work of Bratman [Bratman, 1987]; it involves an agent committing to the goals it has already adopted, and tending to bypass (or "filter out") new options that would conflict with their successful completion [Bratman et a/., 1988; Pollack, 1992; Pollack et a/., 1994]. We and others have studied the effectiveness of filtering in domains with various characteristics[Kinny and Georgeff, 1991; Pollack et a/., 1994]. The original filtering strategy was designed as a method for an individual agent to focus its reasoning in a dynamic, but not necessarily multi-agent, environment. Here, we generalize this strategy to multi-agent environments. Where single-agent filtering means tending to bypass …