Neural Network-Based Position and Orientation Estimation of a Centimeter Scaled Robot Using a Localized Magnetic Field Map

Abstract

This paper develops a position and orientation estimation system for a robot moving over a plane based on the use of an actively controlled magnetic field. The position estimation system consists of two magnetic sensors on the robot and an actively controlled rotating permanent magnet. The orientation of the magnet is controlled to roughly point at the robot and a localized magnetic field map based on a neural network is developed for a narrow region around the pointing direction of the magnet. Using the magnetic fields measured at both sensors, the radial and polar positions and the orientation of the robot are estimated using an unscented Kalman filter. The orientation of the magnet is then more finely controlled to point precisely at one of the magnetic sensors. This enables the further design of an asymptotically stable nonlinear observer that provides enhanced accuracy in the radial position estimation of the robot. Extensive experimental results are presented on the performance of the estimation system, including real-time estimation of both the moving robot’s two-dimensional position and its orientation.