Development of a Salted Extended Kalman Filter for the Hybrid-Dynamic Walking of a Bipedal Robot

Accurate state estimation is critical for the controller design of bipedal robots such as the
Hybrid Zero-Dynamics approach. This work investigates a Salted Extended Kalman Filter
(SEKF), an EKF variant for hybrid dynamics, within a real-time simulation of a planar five-link
robot model. Measurement noise characteristics are identified from encoder data from a real
robot prototype and the SEKF is tested under both actual and augmented noise parameters.
The SEKF demonstrates reliable state estimation across impacts, effective noise suppression,
and facilitates controller performance, and necessary extensions for deployment on a physical
prototype are presented.