Factor graph-based multi-IMU (Inertial Measurement Unit) state estimation method for legged robot

By optimizing the model using a factor graph sliding window and employing adaptive probabilistic scaling zero-speed update constraints, this approach addresses the issues of linearization error accumulation and insufficient robustness to foot slippage and dynamic impacts in existing technologies. It achieves high-precision and robust state estimation for legged robots, enabling autonomous navigation and dynamic motion in complex environments.

CN122448199APending Publication Date: 2026-07-24NAT UNIV OF DEFENSE TECH
0 Cites 0 Cited by

Patent Information

Application Number
CN202610767003.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-05-29
Publication Date
2026-07-24

AI Technical Summary

Technical Problem

Existing EKF ontology-aware state estimation methods based on single or multiple IMUs struggle to effectively handle non-ideal contact situations such as foot slippage in complex environments lacking reliable external perception. Furthermore, they lack robustness to transient nonlinear dynamic disturbances during motion, resulting in limitations in the accuracy and reliability of state estimation.

Method used

A sliding window optimization model based on factor graphs is adopted, which closely integrates data from the main trunk IMU, multiple leg IMUs and joint encoders. Adaptive optimization of state estimation is achieved by adaptive probability scaling zero-rate update constraints and constructing kinematic relative position factors.

Benefits of technology

It improves the accuracy and robustness of state estimation under unstructured terrain and dynamic motion, effectively suppresses the accumulation of linearization error, achieves high-precision whole-body motion state perception, and enhances the system's robustness to single sensor failures or noise.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention relates to the technical field of robot inertial navigation, and discloses a factor graph-based multi-IMU (Inertial Measurement Unit) state estimation method for a foot-type robot, which comprises the following steps of: constructing a factor graph-based sliding window optimization model, distributing a plurality of key frames in time, and enabling each key frame to comprise a trunk inertial state and a plurality of leg inertial states of the foot-type robot; acquiring multi-source sensor data, wherein the multi-source sensor data comprises inertial measurement data of a trunk inertial measurement unit, inertial measurement data of leg inertial measurement units of legs and joint encoder data of leg joints; in the sliding window optimization model, tight coupling constraint factors are established for the multi-source sensor data, the constraint factors comprise an inertia pre-integration factor, a zero-speed updating factor scaled according to probability and a kinematics relative position factor, the sliding window optimization model is solved, maximum posteriori estimation is executed, and the multi-source sensor data is obtained. And outputting the optimized trunk inertia state and leg inertia state of the foot robot in the current sliding window.
Need to check novelty before this filing date? Find Prior Art