2009 Special Issue: Optimal control of unknown affine nonlinear discrete-time systems using offline-trained neural networks with proof of convergence

摘要: The optimal control of linear systems accompanied by quadratic cost functions can be achieved solving the well-known Riccati equation. However, nonlinear discrete-time is a much more challenging task that often requires Hamilton-Jacobi-Bellman (HJB) In recent literature, approximate dynamic programming (ADP) techniques have been widely used to determine or near policies for affine systems. an inherent assumption ADP value controlled system one step ahead and at least partial knowledge dynamics known. this work, need relaxed in development novel approach using two part process: online identification offline training. First, process, neural network (NN) tuned tuning laws learn complete plant so local asymptotic stability error shown. Then, only learned NN model, attempted resulting law. proposed scheme does not require explicit as model needed. proof convergence demonstrated. Simulation results verify theoretical conjecture.