A Dynamic Neural Network with Feedback for Trajectory Generation

A Dynamic Neural Network (DNN), comprising of a Recurrent Neural Network (RNN) and two Feedforward Neural Networks (FFNN), is detailed in this paper. This neurodynamic system is used to adaptively solve the problem of trajectory generation; planning a path connecting a locus of points as a function of time. The BAN is designed to exhibit desired dynamic behavior which drives the output FFNN to generate a specific trajectory. To make the system adaptive, a feedback FFNN is introduced to generate inputs to the R\N based on the outputs of the output FFNN. Simulations conducted for generation of different types of trajectories show the capabilities of the DNN with feedback in solving the trajectory generation problem. (C) 2016, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.