Abstract
Magnetic sensors have been utilized to locate the orientation and position of solitary permanent magnets (PM) through characterization of its unique nonlinear magnetic field. Similar characterization can be applied to the existing symmetric assembly of multiple PMs in the rotors of PM-based spherical actuators. This paper introduces a sensing method that exploits the symmetry of the rotor PM configuration to reduce the required characterization through discretization and classification of the global magnetic flux density mapping into smaller segments and sectors. Unlike existing methods that use fitted analytical models to extract orientation estimates, this method engages a trained neural network.