High Efficiency Data Driven Control Based on Dynamic Linearization and PIDNN With Cohen-Coon for Discrete Nonlinear Fast Time-Varying Systems

In this brief, a high efficiency data driven control algorithm based on dynamic linearization (DL) and proportional-integral-derivative neural network (PIDNN) with Cohen-Coon (CC) is designed to achieve trajectory tracking control of discrete nonlinear fast time-varying systems (DNFTS). Firstly, CC approach is used to reliably obtain initial parameters of PIDNN and initial pseudo partial derivative (PPD) of DNFTS by exciting systems to produce input/output (I/O) data, meanwhile, DL method utilizes system I/O data to online establish a virtual data model equivalent to DNFTS. Secondly, the equivalent virtual data model generates PPD to online compensate PIDNN weights, which guarantees that PIDNN weights can be promptly tuned in the right direction. Moreover, one algorithm complexity criterion is defined to estimate proposed algorithm execution efficiency. The simulation studies indicate that proposed DL-PIDNN-CC has high execution efficiency in terms of fast convergence speed with low computational burden and has superior control performance in terms of Integral Squared Error (ISE) and Integral Absolute Error (IAE).