Multiple-round timber auction design and simulation

This paper presents a multiple-round timber auction simulation, developed in order to study various configurations of auction design. In this study, simultaneous sequential timber auctions are modelled and analysed using agent-based simulation technology. As there are many individual items in the auction to be sold, the auction designer defines several rounds that are sequential at pre-defined intervals. At each round, the auction designer announces several simultaneous auctions. Since bidders are offered different items at each round, a mathematical linear programing model for selecting the best set of items to bid for is presented. Different bidding patterns are simulated and compared in various setup configurations. The most advanced of these strategies are adaptive and use agent-learning capability. The comparisons include the success rate of winning the auction and the winning price per m3. This study suggests an efficient bidding pattern for bidders to bid in order to achieve to their goal and increase their profit. Similarly, in order to increase profit, the auctioneer (i.e. the government) needs to control several auction parameters including the number of auctions per year, the lot size, the auction periodicity, and the number of bidders. This study also suggests parameters configurations that to maximise revenue for the auctioneer.

Mots clés

Timber auction, Sequential auction, Learning strategy, Multi agent system, Allocation