Simple and Effective State Estimators for Humanoid Robots Under Different Noise Conditions

The promise of humanoid robots over standard wheeled robots is to provide improved mobility over rough terrain. As a high-dimensional nonlinear system with multi-links/joints, however, humanoid robot is typically difficult to control during its moving process, and its state estimation is of critical importance. This paper presents two simple and effective state estimation schemes for humanoid robots, of which one is based on the Linear Inverted Pendulum Model (LIPM) combined with a Kalman Filter (KF) and the other utilizing the nonlinear center of mass (CoM) dynamics integrated with a Dual Loop Kalman Filter (DLKF). Experiments are conducted to evaluate their performances under different noise conditions. Results demonstrate that that the LIPM - KF estimator is computationally speed, yet it is less robust to disturbances as compared to DLKF with nonlinear CoM dynamics, while CoM dynamics estimator is more accurate under high noise conditions. This study illustrates the importance of balancing accuracy and computational load to achieve timely and stable locomotion control for humanoid robots.