Designing the Market Game for a Trading Agent Competition

The authors discuss the design and operation of a trading agent competition, focusing on the game structure and some of the key technical issues in running and playing the game.

Learning about other agents in a dynamic multiagent system

We analyze the problem of learning about other agents in a class of dynamic multiagent systems, where performance of the primary agent depends on behavior of the others. We consider an online version of the problem, where agents must learn models of the others in the course of continual interactions. Various levels of recursive models are implemented in a simulated double auction market. Our experiments show learning agents on average outperform non-learning agents who do not use information about others. Among learning agents, those with minimum recursion assumption generally perform better than the agents with more complicated, though often wrong assumptions.

Auction Protocols for Decentralized Scheduling

Decentralized scheduling is the problem of allocating resources to alternative possible uses over time, where competing uses are represented by autonomous agents. Market mechanisms use prices derived through distributing bidding protocols to determine schedules. We investigate the existence of equilibrium prices for some general classes of scheduling problems, the quality of equilibrium solutions, and the behavior of an ascending auction mechanism and bidding protocol. To remedy the potential nonexistence of price equilibria due to complementarities in preference, we introduce additional markets in combinations of basic goods. Finally, we consider direct revelation mechanisms and compare to the market-based approach.

A Parametrization of the Auction Design Space

We present an extensive breakdown of the auction design space that captures the essential similarities and differences of many auction mechanisms in a format more descriptive and useful than simple taxonomies. This parametrization serves as an organizational framework in which to classify work within the field and uncovers parameter combinations corresponding to novel mechanisms. The structured characterization of auction rules can be exploited for the modular design of configurable auction servers. It also facilitates the communication of auction rules to software agents, enabling the automation of flexible market-based negotiation.