Model-free control of dynamical systems with deep reservoir computing

Abstract We propose and demonstrate a nonlinear control method that can be applied to unknown, complex systems where the controller is based on type of artificial neural network known as reservoir computer. In contrast many modern neural-network-based techniques, which are robust system uncertainties but require model nonetheless, our technique requires no prior knowledge thus model-free. Further, approach does not an initial identification step, resulting in relatively simple efficient learning process. Reservoir computers well-suited problem because they small training data sets remarkably low times. By iteratively adding layers controller, precise law identified quickly. With examples both numerical high-speed experimental systems, we capable controlling highly dynamical display deterministic chaos nontrivial target trajectories.