A Market-Oriented Programming Environment and its Application to Distributed Multicommodity Flow Problems

Market price systems constitute a well understood class of mechanisms that under certain conditions provide e ective decentralization of decision making with minimal com munication overhead In a market oriented programming approach to distributed problem solving we derive the activities and resource allocations for a set of computational agents by computing the competitive equilibrium of an arti cial economy Walras provides basic constructs for de ning computational market structures and protocols for deriving their corresponding price equilibria In a particular realization of this approach for a form of multicommodity ow problem we see that careful construction of the decision process ac cording to economic principles can lead to e cient distributed resource allocation and that the behavior of the system can be meaningfully analyzed in economic terms Distributed Planning and Economics In a distributed or multiagent planning system the plan for the system as a whole is a com posite of plans produced by its constituent agents These plans may interact signi cantly in both the resources required by each of the agents activities preconditions and the prod ucts resulting from these activities postconditions Despite these interactions it is often advantageous or necessary to distribute the planning process because agents are separated geographically have di erent information possess distinct capabilities or authority or have been designed and implemented separately In any case because each agent has limited competence and awareness of the decisions produced by others some sort of coordination is required to maximize the performance of the overall system However allocating resources via central control or extensive communication is deemed infeasible as it violates whatever constraints dictated distribution of the planning task in the rst place The task facing the designer of a distributed planning system is to de ne a computa tionally e cient coordination mechanism and its realization for a collection of agents The agent con guration may be given or may itself be a design parameter By the term agent I refer to a module that acts within the mechanism according to its own knowledge and interests The capabilities of the agents and their organization in an overall decision making structure determine the behavior of the system as a whole Because it concerns the collec tive behavior of self interested decision makers the design of this decentralized structure is fundamentally an exercise in economics or incentive engineering The problem of developing architectures for distributed planning ts within the framework of mechanism design Hur wicz Reiter and many ideas and results from economics are directly applicable In particular the class of mechanisms based on price systems and competition has been deeply investigated by economists who have characterized the conditions for its e ciency c AI Access Foundation and Morgan Kaufmann Publishers All rights reserved