Conference Papers

KAIST Exoskeleton Lab

2025 State Estimation of a Powered Exoskeleton with Dynamic Motion Using Kinematic and Inertial Odometry

본문

Author
Jimin Youn, Jinsu Park, Jongwon Kim, Hyeongjun Kim, Kyoungchul Kong
Conference
10th IFAC Symposium on Mechatronic Systems & 14th IFAC Symposium on Robotics
Volume
59
Number
18
Page
229-234
Year of publication
2025

Abstract:

This paper introduces a state estimation framework for exoskeletons with dynamic motion that combines kinematic odometry with inertial data from IMUs placed on the trunk and shanks. Unlike traditional approaches suited to quadrupedal and humanoid robots, where point-contact models and lightweight foot structures simplify forward kinematics, exoskeletons encounter significant frame or joint deformation and complex heel-to-toe contact patterns. The proposed estimator mitigates the corruption from the deformation and motion acceleration by dynamically adjusting the weighting between kinematic and IMU odometry based on the acceleration of each segment as well as torque applied to each leg. Furthermore, this study proposes a novel quaternion-based time-varying complementary filter (TVCF) structure for individual IMUs to achieve accurate orientation estimation even during dynamic motions. Experimental validation demonstrates that the framework reliably captures state information across dynamically demanding exoskeleton motions, underscoring its efficacy for real-world applications.