An algorithm to reduce human–robot interface compliance errors in posture estimation in wearable robots

Wearable Technologies, 2022 · DOI: 10.1017/wtc.2022.29 · Published: December 1, 2022

Assistive Technology Bioinformatics Telehealth & Digital Health

Simple Explanation

Wearable robots assist users by applying forces, which are determined by controllers that need accurate posture estimation. However, the flexible connection between the robot and the person can cause errors in posture estimation. This study introduces an algorithm that uses machine learning to correct posture estimation errors caused by the compliant interface. Data was collected from participants walking on a treadmill while wearing a wearable robot, and this data was used to train a model to correct for mechanical compliance errors. The algorithm improved the accuracy of thigh angle estimation, which is important for the robot's controller. The results suggest that machine learning can be effectively combined with wearable robot sensors to improve posture estimation.

Study Duration

Not specified

Participants

8 (4 females)

Evidence Level

Not specified

Key Findings

1
The algorithm reduced the estimated thigh segment’s angle RMS error from 6.3° to 2.5°.
2
The average maximum error in thigh angle estimation was reduced from 13.1° to 5.9°.
3
Improvements in posture estimation were observed across various assistance force levels and walking speeds.

Research Summary

The study addresses the problem of posture estimation errors in wearable robots due to the compliance of the human-robot interface. A novel algorithm is presented that uses machine learning to correct these errors. The algorithm was trained using data from participants walking on a treadmill while wearing a wearable robot, the Myosuit. The results showed a significant reduction in the RMS and maximum errors in thigh angle estimation. The findings suggest that machine learning algorithms can be effectively used with wearable robot sensors to improve user posture estimation, leading to better control and assistance delivery.

Practical Implications

Improved Wearable Robot Control

More accurate posture estimation leads to better control of wearable robots, resulting in more effective and personalized assistance.

Enhanced User Comfort

Reducing posture estimation errors can prevent the robot from applying forces inappropriately, increasing user comfort.

Real-Time Implementation

The algorithm can be implemented in real-time, making it suitable for use in dynamic and unpredictable environments.

Study Limitations

1
The relative motion between the robot and the human was quantified using camera-based motion capture systems.
2
Only constant walking speed profiles of 0.8 and 1.3 m/s, precisely controlled by using a constant-speed treadmill, were used.
3
The fitting of the robot on the participants was neither controlled nor measured during the donning procedure.

Your Feedback

Was this summary helpful?

Back to Assistive Technology