摘要
针对具有通信约束的量化控制系统辨识过程,采用伯努利随机重复性试验方法产生辨识数据,分析了在随机重复性试验方法下的量化辨识数据数字特征和持续激励性,并将多划分区间的量化辨识数据筛选问题建模为二值(0或1)优化问题;接着放松二值约束条件,将二值优化问题转化为凸优化问题求解,同时给出了近似优化问题的迭代算法;最后通过数字仿真验证了方法的有效性.
This paper studies the selection of quantized identification data for control systems subjected to communication constraints.A repeated stochastic Bernoulli experiments method is employed to generate identification measurement data,then,the mathematic characteristics and the persistent exciting condition of the quantized identification data are evaluated.The selection of quantized identification data problem is so formulated as a two-way optimization problem when a set of quantization partitions are given,with the aim to minimize the error of estimation.After that,an approximate reformulation of the two-way partition optimization problem is presented by relaxing the two-way partition constraints of the optimization variables.The solution of the problem is obtained heuristically by interior-point algorithm(Newton's method).Simulation results show the effectiveness of the conclusions.
出处
《华中科技大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2009年第S1期234-237,共4页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
国家自然科学基金资助项目(60804013)
关键词
系统辨识
控制系统
凸优化
信号量化
随机试验
内点法
system identification
control system
convex optimization
signal quantization
stochastic experiment
interior algorithm
