Solving a reverse auction problem by bi-level distributed programming and genetic algorithm

As globalisation and international sourcing prevail, constructing an optimal combination of diverse suppliers has become of great importance for meeting measurable objectives, such as on-time delivery, cost efficiency and risk mitigation. This paper presents a method for solving a sealed-bid, multi-issue, multi-sourcing reverse auction problem, where a buyer distributes his demand to multiple suppliers and each supplier responds by submitting a bid price to the buyer. The problem is formulated as a bi-level distributed programming model in which the buyer is an upper-level decision-maker, while suppliers at a lower level make decisions independently to each other. The negotiation process is facilitated via iterative exchanges of decision information between the buyer and suppliers. We used a genetic algorithm to establish an optimum quantity allocation at the upper level. In the lower level decision-making process, the concepts of revenue management are employed to coordinate pricing and (production) scheduling decisions. We also conducted three groups of simulation experiments to assess the quality of the proposed solution, as well as to examine its computational efficiency under various parameter settings. The results were consistent with the expectation.