Attention-Based Neural-Augmented Kalman Filter for Legged Robot State Estimation

In this letter, we propose an Attention-Based Neural-Augmented Kalman Filter (AttenNKF) for state estimation in legged robots. Foot slip is a major source of estimation error: when slip occurs, kinematic measurements violate the no-slip assumption and inject bias during the update step. Our objective is to estimate this slip-induced error and compensate for it. To this end, we augment an Invariant Extended Kalman Filter (InEKF) with a neural compensator that uses an attention mechanism to infer error conditioned on foot-slip severity and then applies this estimate as a post-update compensation to the InEKF state (i.e., after the filter update). The compensator is trained in a latent space, which aims to reduce sensitivity to raw input scales and encourages structured slip-conditioned compensations, while preserving the InEKF recursion. Experiments demonstrate improved performance compared to existing legged-robot state estimators, particularly under slip-prone conditions.

💡 Research Summary

This paper introduces the Attention‑Based Neural‑Augmented Kalman Filter (AttenNKF), a novel state‑estimation framework for legged robots that explicitly addresses foot‑slip‑induced bias. Traditional proprioceptive estimators, such as the Invariant Extended Kalman Filter (InEKF), assume no slip at contact points; when slip occurs, the kinematic update injects a systematic error that is usually mitigated only indirectly by inflating measurement covariance or discarding updates. AttenNKF departs from this paradigm by (1) defining a continuous slip level for each foot, derived from a CNN‑based contact classifier and the magnitude of forward‑kinematics‑computed foot velocity, mapped through a sigmoid to a value in