Automated Negotiation in Many-to-Many Markets for Imperfectly Substitutable Goods

Abstract

Abstract. In this paper, we present an agent which is able to negotiate the buying and selling of imperfectly sustitutable goods in a double auction style market. Two goods are said to be imperfectly substitutable if a buyer can use either of them, but prefers one over the other. For example, an electronics manufacturer using a RAM chip can use many suppliers to do this but may be willing to pay a premium for components with a lower failure rate. We give a formal description of a double-auction style market mechanism for trading such goods, and define the (classical) equilibrium in such an environment. We present the IS-ZIP agent, which is a generalisation of the ZIP agent for double auctions of Cliff and Bruten [3]. It is able to participate in our double auction environment to make purchases or sales of imperfectly substitutable goods. We demonstrate that, when trading a single good, it is equivalent to Preist and van Tol's modification of the ZIP agent [11]. We describe experiments where a group of IS-ZIP agents with different valuations trade repeatedly, and demonstrate that they rapidly converge to our predicted equilibrium. We conclude by relating our work to that of others, particulary work dealing with multi-attribute negotiation, and discussing extensions.

Agent-based service composition through simultaneous negotiation in forward and reverse auctions

“…These include heuristic approaches ( [4], [21]), stochastic analysis ( [14], [15]), fuzzy logic [9] and dynamic programming [28]. Work has also been carried out to distribute [13] and generalise [20] the environment.…”