Neural Network Based Three Axis Satellite Attitude Control Using Only Magnetic Torquers

Abstract

Magnetic actuation utilizes the mechanic torque that is the result of interaction of the current in a coil with an external magnetic field. A main obstacle is, however, that torques can only be produced perpendicular to the magnetic field. In addition, there is uncertainty in the Earth magnetic field models due to the complicated dynamic nature of the field. Also, the magnetic hardware and the spacecraft can interact, causing both to behave in undesirable ways. This actuation principle has been a topic of research since earliest satellites were launched. Earlier magnetic control has been applied for nutation damping for gravity gradient stabilized satellites, and for velocity decrease for satellites without appendages. The three axes of a micro-satellite can be stabilized by using an electromagnetic actuator, which is rigidly mounted on the structure of the satellite. The actuator consists of three mutually orthogonal air-cored coils on the skin of the satellite. The coils are excited so that the orbital frame magnetic field and body frame magnetic field coincides i.e. to make the Euler angles to zero. This can be done using a Neural Network controller trained by PD controller data and driven by the difference between the orbital and body frame magnetic fields.