Planar Pose Estimation System Design via Explicit Spatial Representation Model of Concurrent AC Magnetic Fields

Abstract

This study proposes a new approach to pose estimation algorithm and system implementation based on concurrent AC magnetic fields from an array of magnetic coils. The typical dipole model used in the existing research has inappropriate prediction characteristics when the operating range of dynamic objects is widened, as in the cases of mobile robots or drones. An enlarged circular loop with a meter-sized coil dimension and its corresponding excitation electronics are developed to overcome this shortcoming. In designing a pose estimation algorithm, a compact mathematical model is derived through the suggested spatial representation and the partial derivatives with respect to states. As a result, an accurate and computationally efficient pose estimation framework has been achieved, which validates the feasibility of the developed spatial representation model with actual magnetic measurements. While the proposed method can provide a general formula for a 3D dynamic case, yet a reduced model is specifically presented for a typical indoor 2D application. The performance of the proposed algorithm is demonstrated through the ground experiment in an indoor building environment with 50m2 coverage. Compared with the ground truth, static error variation from the low-order polynomial case is computed to be sub-centimeter, which allows an onboard algorithm implementation. Finally, the experiment via dynamic trajectory tests demonstrated the relative position accuracy of 1.2 % and heading error of about 2 degrees within the operating range.