Session: 06-14-01: Biotechnology and General Applications

Paper Number: 116592

116592 - Upper Body Joint Angle Calculation and Analysis Using Multiple Inertial Measurement Units 

Biomechanics could play a crucial role in finding solutions to issues like work-related musculoskeletal disorders (WMSDs) that are common in professions that require unnatural postures, such as commercial fishing and farming. Understanding how workers in these professions move can help find solutions to WMSDs. While motion capture systems are widely used for body posture measurement, their lack of portability limits their use in remote environments. To address this issue, IMU-based measurement could provide a potential alternative. An IMU (Inertial Measurement Unit) consists of a triaxial accelerometer, gyroscope, and magnetometer, which can be used to build an estimation of the sensor orientation with respect to the world frame of reference. In principle, the pair of the accelerometer and the magnetometer can be used to estimate orientations, or the gyroscope data of angular velocity alone can be numerically integrated to estimate orientations. However, these sensors are susceptible to gyroscope drift due to the integration process, accelerometer noise in dynamic conditions, and magnetic hard and soft iron distortions.

In this paper, three sensor fusion algorithms previously developed to account for those issues while combining the benefits associated with each sensor were applied to estimate joint angles of the shoulder and arm effectively: Madgwick's filter, Kalman filter, and complementary filter. Madgwick's filter utilizes the gradient descent algorithm to quickly converge toward an orientation estimation. Kalman filters are an iterative two-step process involving a prediction followed by an update, using knowledge of measurement uncertainties and their Gaussian distributions in each iteration. Complementary filters use a weighted average to combine the sensor data one at a time and find an angle and axis of rotation to tilt the orientation closer to the actual value. In the implementation of the algorithms, quaternions are used as an ideal orientation representation in order to avoid singularities throughout the process.

To demonstrate the application of the three algorithms for arm motion tracking, three IMUs were placed on each part of an arm (lower, upper, and shoulder) for two representative motions: reaching up and reaching across a desk. Data from the three IMUs was taken simultaneously and then post-processed to estimate the elbow and shoulder joint angles. These estimations were compared with the reference data collected using a motion capture system to determine the overall accuracy for each sensor fusion method. Furthermore, parameters for each sensor fusion algorithm were optimized to minimize the overall joint angle errors. Future work includes exploring other sensor fusion approaches such as extended or unscented Kalman filters and using multiple IMUs per joint to improve accuracy.
 

Presenting Author: Ji-Chul Ryu Northern Illinois University

Presenting Author Biography: Dr. Ji-Chul Ryu received the B.S. and M.S. degrees in mechanical engineering from Korea Advanced Institute of Science and Technology (KAIST) and the Ph.D. degree in mechanical engineering from the University of Delaware in 2009. He is currently an Associate Professor of the Mechanical Engineering Department at Northern Illinois University. From 1999 to 2004, he was an engineer at several companies, such as Samsung, where he developed various types of automated robotic machines. He worked as a Postdoctoral Fellow with the Neuroscience and Robotics Laboratory, Northwestern University prior to joining Northern Illinois University in 2013. He's been a member of editorial board of Advances in Robotics Research. Also, he’s a member of ASME (American Society of Mechanical Engineers), affiliated with Dynamic Systems & Control and of IEEE (Institute of Electrical and Electronics Engineers), affiliated with Robotics and Automation Society. His research interests include dynamic nonprehensile robotic manipulation, intelligent human-robot interface for work collaboration,
wearable sensor-based state estimation for biomechanical studies, and integrated planning and control of autonomous robot manipulators.

Authors:

Aaron Freedkin Northern Illinois University
Ji-Chul Ryu Northern Illinois University
Jaejin Hwang Northern Illinois University