Identification of non-linear dynamic model of UUV based on ESN neural network
Xinqian Bian, Chunhui Mou · Chinese Control Conference · 2011
Unmanned underwater vehicle (UUV) is a highly complex nonlinear dynamic system, and neural network has the ability to arbitrary approximate nonlinear system in theoretically. Furthermore, echo state network (ESN) is a new type recurrent neural network based on state reservoir. To improve the accuracy of UUV's dynamic model, this paper based on the use of echo state networks (ESN) of the system identification method, using “meta-learning” strategy for offline training ESN network and genetic algorithm to optimize the main parameters, to remove the difficulty of choosing the ESN parameters. This method was applied to approximate of dynamic model of six degree of freedom of UUV, and build on the dynamic model. Finally, the simulation proved that the network structure identification algorithm has a good approximation ability and fast training speed.