Auction in dynamic environments: incorporating the cost caused by re-allocation

• Authors: Shigeo Matsubara
• Venue: Proceedings of the fourth international joint conference on Autonomous agents and multiagent systems
• Year: 2005

This paper proposes an auction protocol for solving a resource allocation problem in dynamic environments. In such environments, the valuation of resources has uncertainty for each bidder, i.e., this valuation depends on the situation not only at the point when the auction is held but also at the point when the allocated resources are actually used. For example a bidder's valuation in fine weather may be different from that in rainy weather. A solution for dealing with this problem is to execute auctions whenever an event occurs and then to re-allocate resources. Re-allocating resources, however, may cause disutility. Moreover, it does not always provide an equilibrium strategy because it can be viewed as a sequential auction, which means that we cannot accurately predict what outcome will be obtained. To solve this problem, we propose an auction protocol that allows bidders to declare the cost due to re-allocation and then decides an allocation based on this cost of re-allocation as well as the surplus obtained from the allocated resources themselves in the realized situation. We prove that a bidder's truth telling is in equilibrium and that a socially efficient allocation is obtained in the proposed protocol.