Study On Model And Optimization Algorithm Of Bidding Based On Game Theory

With the market-oriented development of domestic project purchasing, more and more contractors are selected through bidding. When the enterprise cannot complete a large scale project separately, it has to seek partners through the tender, and it is called enterprises dynamic alliance. The modern dynamic alliance enterprises are forming a three-dimensional network of a large scale alliance. The sponsor often requests the members to form a combination set based on their different core competencies and positions. The optimization partner combinations of dynamic alliance have received intensive attention.Because of the limitations of resources and profits space, there is not only competition but also coordination between the sponsor and members, or between the contractors of difference subprojects. Thus, an optimization decision is reached to the goal of win-win. On the basis of analyzing the optimization goal of subprojects, using the game theory, this thesis developed a nonlinear integer-programming model of combined enterprises partner selection. The evolutionary game algorithm is designed to solve the problem. The proposed algorithm sets up relevant mappings from the partner combinations to the players' strategy combinations, from the sponsor objective function to the players' utility function, and the optimization partner combinations to the Pareto dominance Nash equilibriums. The results of simulations suggest that the proposed method is effective in both running time and performance, in partner selection of enterprise alliance under large scale random conditions. At last the thesis improves on the evolutionary game algorithm using multi-agent reinforcement learning. A case study was carried out to compare the proposed algorithm with genetic algorithm and evolutionary game algorithm in the. The simulation results of different scales suggest that the proposed method is more effective in partner selection of enterprise alliance under larger scale random conditions.