Foresight-based pricing algorithms
in an economy of software agents
Gerald J. Tesauro and Jeffrey O. Kephart
IBM T. J. Watson Research Center
30 Saw Mill River Rd., Hawthorne NY, 10532
Abstract:
We propose several heuristic approaches to the
development of pricing algorithms for software agents that
incorporate foresight, i.e., an ability to model and predict
responses by competitors. In the absence of foresight,
prior work has shown that, in an economy
of myopic software agents, undesirable system behaviors such
as endless price wars can frequently occur
(Kephart et al., 1998). We show how the
introduction of even the smallest amount of lookahead in the agents'
pricing algorithms can significantly reduce or eliminate the
occurrence of price wars. We also investigate two approaches to
developing algorithms that are capable of deep lookahead, while
avoiding the classic problem of infinite recursion of opponent
models. The two approaches are based on adaptations of: (i) the
classic minimax fixed-depth search algorithms used in two-player
games such as chess; (ii) dynamic programming (DP)-style algorithms,
that have recently been extended to the domain of two-player
zero-sum Markov games (Littman, 1994).
|
|
