研究期間：10108~10207;This proposal will analyze the variation characteristic of multiple points seeking method in solution space by means of signal processing, and design a filter to improve the multiple points seeking method so that the algorithm can obtain the globally optimal solution. The proposal is divided into two directions： (1)How to apply the signal processing to analyze the characteristic of multiple points seeking method under various frequencies. (2)How to design a filter to improve the multiple points method to enhance its efficacy. The proposal will present various assumptions and try to get the best fit algorithm aimed at achieving the faster and the better online requirement. As is well known, there are currently some methods, such as the GA method or the particle swarm optimization method in which provided that the variation property between the individual and the group has a certain specific spectrum, then employ a filter to adjust the variation property to fast search for the optimal solution of the problem. However, whenever the multiple points method iterates, the entire individuals will be altered. This implies it will generate more sets of signal. If the number of individuals is large, the number of the signal is increased simultaneously. Therefore, how to find an algorithm to combine the both techniques will be a great topic to break through. This proposal will combine the technique of signal processing and the multiple points method to improve the problem that the method is not easy to be applied under the online circumstances. The results will be validated in a general functional solution, and then will be applied in a practical system, also compared with the conventional method, in order to achieve the target that is better and faster.
