Satellite attitude control through evolving a neural network

We propose a pure topological recurrent network controller for satellite attitude control, which has random binary connections in hidden layer, and all hidden neurons are activated by sinusoidal functions. A direct graph encoding method and four genetic operators are implemented for using genetic programming to train this controller. Moreover, a simulated small satellite which equipped with three reaction wheels was developed, then this simulator was employed to test the controller and training method for a given simple attitude adjusting mission. The experimental results reveal that this controller has the simplicity, usability and potentials for satellite attitude control through evolutionary learning.