Neural Network Control of Space Vehicle Intercept and Rendezvous Maneuvers

Neural networks are examined for use as optimal controllers. The effect of the addition of noise to the neural network input measurements is investigated to determine the performance robustness of the neural network controllers. These techniques are applied to the autonomous control of interceptor-to-target rendezvous missions. For this example, the target lies in a circular orbit and remains passive throughout the maneuver. The linearized Clohessy–Wiltshire equations with thrust are used to describe the relative motion of the two vehicles. Parameter optimization is used to generate the training data for the neural network designs. A combination of open-loop and closed-loop control is shown to work effectively for this problem.