A partheno-genetic algorithm for optimal winner determination in combinatorial auctions

Combinatorial auction is an efficient mechanism for allocating items among agents. And winner determination is the core problem in combinatorial auction. It is well known that the problem of determining the winners, so as to maximize revenue, is a NP-complete problem. In this paper, two fully expressive bidding languages are introduced firstly, namely XOR-bids and OR-bids, with which bidders can place additive or exclusive bids over collection of combinations. Meanwhile, some heuristic rules about the bidding languages are presented. Then with partheno-genetic operators and the fitting-first heuristic rules, a partheno-genetic algorithm is presented on the basis of the ideas of heuristic algorithm. The algorithm is simple for implementing and can achieve fine outcome to solve the winner determination problem without requiring complex operators of crossover and mutation. Simulation results show that the algorithm has very high feasibility and efficiency.