Rapid dynamical pattern recognition for sampling sequences
Rapid dynamical pattern recognition for sampling sequences作者机构:School of Automation Science and Engineering South China University of Technology School of Control Science and Engineering Shandong University Department of Mechanical Industrial and Systems Engineering University of Rhode Island
出 版 物:《Science China(Information Sciences)》 (中国科学:信息科学(英文版))
年 卷 期:2021年第64卷第3期
页 面:118-135页
核心收录:
学科分类:0711[理学-系统科学] 0810[工学-信息与通信工程] 0808[工学-电气工程] 07[理学] 071101[理学-系统理论] 0812[工学-计算机科学与技术(可授工学、理学学位)]
基 金:supported in part by National Natural Science Foundation of China (Grant No. 61890922) in part by National Major Scientific Instruments Development Project (Grant No. 61527811)
主 题:deterministic learning dynamical pattern recognition sampling sequence synchronization
摘 要:In this paper, based on deterministic learning, we propose a method for rapid recognition of dynamical patterns consisting of sampling sequences. First, for the sequences yielded by sampling a periodic or recurrent trajectory(a dynamical pattern) generated from a nonlinear dynamical system, a sampled-data deterministic learning algorithm is employed for modeling/identification of inherent system dynamics. Second,a definition is formulated to characterize similarities between sampling sequences(dynamical patterns) based on differences in the system dynamics. Third, by constructing a set of discrete-time dynamical estimators based on the learned knowledge, similarities between the test and training patterns are measured by using the average L1 norms of synchronization errors, and general conditions for accurate and rapid recognition of dynamical patterns are given in a sampled-data framework. Finally, numerical examples are discussed to illustrate the effectiveness of the proposed method. We demonstrate that not only a test pattern can be rapidly recognized corresponding to a similar training pattern, but also the proposed recognition conditions can be verified step by step based on historical sampling data. This makes a distinction compared with the previous work on rapid dynamical pattern recognition for continuous-time nonlinear systems, in which the recognition conditions are difficult to be verified by using continuous-time signals.