Simulation of three-rate neural network control for stochastic model of a fighter aircraft
Igor Astrov · 2004
The nature of the multirate dynamics of the process makes it very attractive for applications, since the multirate phenomena are complex. This paper presents a research methodology for describing a three-rate continuous-time stochastic control system as state-space (SS) type decentralized models of multiinput/multioutput (MIMO) stochastic control subsystems with the three neural networks (NNs) of different structure. The block diagram of the decentralized control subsystems with the radial basis control NNs for the "fast" and "intermediate" subsystems, and with one linear layer control NN for the "slow" subsystem is designed. An illustrative design example - NN control of an experimental fighter aircraft model in the short approach to landing task - was carried out using the proposed three-rate SS decentralization technique. The simulation results with use of software package Simulink demonstrate that this research technique would work for real-time MIMO stochastic systems.