Moving Mass Attitude Law Based on Neural Networks

Missile designers are becoming increasingly concerned with cost and cost-effectiveness. New techniques such as Moving-Mass Control Systems are being explored for their apparent cost advantage. This paper investigates the ability of a moving-mass attitude control system to control a vehicle with three-axis stabilization in intra-atmospheric space. The general nonlinear equations of motion with three internal moving masses are used to describe the coupling influence to the system caused by the relative movement of the moving masses to the vehicle's shell. The rapid self-learning and adaptive capacities of radial basis function (RBF) neural networks were exploited to revise the Proportional Integral Differential (PID) controller to calculate the positions of the moving masses. The optimal solution is determined by optimizing the weights of the network through Genetic Algorithm (GA) training. With the coordination of the control, the masses are positioned independently to generate modest attitude corrections for the vehicle. Simulation results show the method to be effective in system control as static system stability is achieved to optimally adjust the missile attitude.