Near-Joint Dual-IMU Angle Estimation for Knee Exoskeletons with Adaptive Complementary Filtering

Abstract

Inertial measurement units provide a low-cost sensing solution for knee exoskeleton joint angle estimation, but near-joint IMUs are susceptible to dynamic acceleration induced by link vibration and motion impacts. This paper proposes an adaptive Mahony-type complementary filtering method for sagittal-plane knee angle estimation using dual near-joint IMUs. The method combines dual-IMU relative angular velocity integration with acceleration projection angle feedback. To improve robustness, Gaussian confidence weights based on dual-IMU acceleration-magnitude deviation are introduced to adaptively regulate the proportional-integral (PI) feedback gains. Experiments on a knee exoskeleton platform show that the proposed method achieves a root mean square error (RMSE) of 2.41° and a mean absolute error (MAE) of 1.87° with respect to the encoder reference. Compared with the fixed-gain Mahony PI filter, the proposed method reduces the RMSE and MAE by approximately 29.1% and 32.0%, respectively, demonstrating improved robustness under dynamic motion disturbances.