Residual-Based Adaptive Kalman Filtering for Legged Robot State Estimation

2026-08-03

AI Summary

This work addresses the degraded state estimation accuracy of legged robots under diverse gaits and environments caused by fixed noise parameters in invariant extended Kalman filters (InEKF). To overcome this limitation, we propose an online adaptive strategy that dynamically tunes the observation noise covariance based on filter residuals and innovations. Relying solely on IMU and leg kinematics—without requiring foot contact force sensors or manual parameter tuning—the method significantly enhances estimation robustness and accuracy. Experimental validation on a Unitree Go2 quadruped robot in both indoor and outdoor settings demonstrates a 25% reduction in position estimation error during trotting compared to the standard InEKF with fixed parameters, achieving performance comparable to approaches that leverage foot force measurements.

Abstract

State estimation is a key component in model-based control of walking robots and, more broadly, applicable wherever hidden variables must be inferred. The Kalman filter is widely used to estimate floating-base position and velocity by fusing multiple sensing modalities. However, tuning noise parameters is challenging and typically requires expert knowledge. Moreover, fixed noise parameters are unsuitable for varying gaits and environments. We propose an online adaptation strategy for the process noise covariance matrix Q and the measurement noise covariance matrix R. Specifically, we introduce a filter residual and innovation-based covariance adaptation method for legged robot state estimation and evaluate it against a baseline approach relying on IMU and foot force measurements. The proposed adaptation is implemented within an Invariant Extended Kalman Filter (InEKF) fusing IMU and leg kinematics. Experiments on indoor and outdoor datasets with a Unitree Go2 quadruped show that adapting R is sufficient and improves accuracy by 25% for the trotting gait compared to the fixed-tuned InEKF. Finally, the proposed residual-based adaptation achieves comparable performance to the foot force approach, without requiring foot force measurements or additional parameter tuning.

Problem

• state estimation
• Kalman filter
• noise covariance
• legged robot
• parameter tuning

Innovation

• Residual-based adaptation
• Invariant Extended Kalman Filter
• Noise covariance tuning
• Legged robot state estimation
• Online adaptation