Design of a sensing system for a spherical motor based on Hall Effect sensors and neural networks

This paper proposes a sensing system to measure 3 rotational angles of a spherical wheel motor (SWM). Unlike conventional motors capable of controlling a single DOF motion only, a SWM is able to provide 3-DOF rotational motions. However, it is challenging to measure the three highly-coupled rotational motions in real time. Unlike some previous sensing systems using optical encoders to measure rotation along each axis separately, a contact-less sensing system such as one composed of Hall Effect sensors is preferred, so as to avoid friction and additional moment inertia, which may damage dynamic performance. In this paper, a sensing system based on a combination of magnetic sensors is proposed, and neural networks are applied to compute rotational angles from measured magnetic field. The paper is organized as followings: distributed multi-pole model (DMP) to obtain the SWM magnetic field distribution (MFD) is demonstrated first; important factors affecting measuring accuracy is researched by simulation then; experimental investigations for a SWM rotor are presented; finally, possible methods to improve proposed sensing system are proposed.