极值搜索算法的研究与进展

Research and development of extremum seeking algorithm

摘要:

回顾了极值搜索算法的发展历程,分析了单变量和多变量极值搜索算法、滑模极值搜索算法、斜率搜索算法等算法的优缺点,着重描述了变参数滑模极值搜索算法和退火回归神经网络极值搜索算法等,并且通过重构系统的损失函数方法将极值搜索算法的应用推广至一般系统,拓展了其应用范围。通过其应用实例说明,极值搜索算法能够解决航空控制、工业生产中的一些关键技术问题,因而有必要深入研究极值搜索算法。

Abstract:

The development of extremum seeking algorithm (ESA) is simply reviewed. Multivariable ESA, Merit and demerit of single variable ESA, ESA with sliding mode and slope seeking algorithm are analyzed. ESA with parameter-varying sliding mode and ESA based on annealing recurrent neural network et al is researched in detail. An approach of structuring the cost function of systems can get over the limits of ESA and extend the application of ESA to general systems. Research shows because of the application of extremum seeking algorithm, some critical technical problems in industry and military have been solved and have more advantages over other optimal algorithms. Therefore, extremum seeking algorithm has good prospects of application.