Adaptive neural control of the deployment procedure for tether-assisted re-entry

An adaptive neural control concept for the deployment of a tethered re-entry capsule is presented. The control concept applies an indirect neural controller that combines two neural networks, a controller network and a plant model network. While the controller network is initialized by means of multiple conventional linear quadratic regulator designs, the plant model network is trained to predict future states, thus deviations from an optimized reference path. inputs are found by means of an online optimization process, which minimizes a user defined cost function, that influences the performance of the neural controller. Due to the special structure of the controller network, stability investigations of the closed control loop become possible. introducing the tether deployment scenario, assumptions and simplifications are applied to the mathematical system model. The numerical simulations focus on the effects of perturbations concerning the initial states and the plant model. The simulation results allow a performance comparison of the linear quadratic regulator and the neural control concept.