A Low-Cost Position Sensing Solution for Miniature Joints in Biomimetic Robots

This paper presents a low-cost and high-precision position sensing solution for biomimetic robotic joints. The system measures the magnetic field of a radially magnetized ring using two orthogonal linear Hall sensors and performs complete angle computation on an embedded MCU. The processing chain includes signal calibration, arctangent decoding, adaptive filtering, and look-up table compensation. An elliptical correction model compensates sensor offset, amplitude imbalance, and phase error. A hybrid Kalman-EWMA filter is applied with adaptive switching to balance dynamic response and noise suppression under varying operating conditions. A bidirectional interpolation compensation method, calibrated with a high-precision reference encoder, further corrects nonlinear errors from assembly and magnetic interference. Implemented on a 180 MHz MCU using its integrated dual 12-bit ADCs for acquisition and real-time processing, the system achieves an angular error within ±0.5° and supports 10kHz output with less than 15 μs delay. The proposed solution offers an effective balance among accuracy, response speed, and cost, demonstrating robustness, reliability, and practical suitability for compact biomimetic joint applications.