Research on a Robust High-Precision Proprioception-Centric State Estimation System for Humanoid Robots

Key Points

• The research aims to develop a robust, high-precision state estimation system for humanoid robots that can handle complex environments.
• Designed a proprioception-centric sensor fusion framework.
• Incorporated kinematic and dynamic constraints to address disturbance from ground impacts.
• Established a mechanism to correct and constrain proprioceptive data with external observations.
• Enhanced external perception using high-frequency proprioceptive precursors.
• Achieved up to 92% reduction in relative translation error compared to existing advanced algorithms.
• Maintained stable estimation capability in extreme degraded scenarios with scarce features.
• Demonstrated high robustness against common sensor failures in complex environments.