Real-time optimal control for irregular asteroid landings using deep neural networks

Abstract

To improve the autonomy and intelligence of asteroid landing control, a real-time optimal control approach is proposed using deep neural networks (DNN) to achieve precise and robust soft landings on asteroids with irregular gravitational fields. First, to reduce the time consumption of gravity calculation, DNNs are used to approximate the irregular gravitational fields of asteroids based on the samples calculated by a polyhedral method. Second, an approximate indirect method is presented to solve the time-optimal landing problems with high computational efficiency by taking advantage of the trained DNN-based gravity model and a homotopic technique. Then, five DNNs are developed to learn the functional relationship between the state and optimal actions obtained by the approximate indirect method, The resulting DNN-based landing controller can generate the optimal control instructions according to the flight state and achieve the real-time optimal control for asteroid landings. Finally, simulation results of the time-optimal landings for Eros are given to substantiate the effectiveness of these techniques and illustrate the real-time performance, control optimality, and robustness of the developed DNN-based optimal landing controller.