Recent research progress of Prof. Fu’s team in human-augmented robotics area - News & Events - SUSTC Department of Mechanical and Energy Engineering
Release date：2020-06-11 views：
Recently, Professor Chenglong Fu (MEE, SUSTech) led his team to make important progress in the human-augmented robotics area. Their papers were published on several well-known academic journals: IEEE Transactions on Cybernetics、IEEE Transactions on Neural Systems and Rehabilitation Engineering、Mechanism and Machine Theory、ASME Journal of Mechanisms and Robotics.
The paper published on IEEE Transactions on Cybernetics reported their research about intelligent prosthesis with an “eye”.
Amputees with powered prosthesis walking in complex environments is a typical “human-prosthesis-environment” coordination problem, which is always challenging. To solve this problem, previous researchers primarily focused on the information interface between humans and prosthesis. Since amputees cannot transmit the neural signals to the prosthesis directly, researchers seek other alternatives. For instance, researchers utilized an inertial measurement unit (IMU) and surface electromyography (EMG) recorded from the residual limb to recognize the motion intent of the amputees. Nevertheless, human signals are usually user-dependent. Besides, human biological and kinetic signals are usually noisy and slow. It is intractable to reconstruct a reliable and timely vision-locomotion loop based on the human–prosthesis interface.
Considering that there were many difficulties to reconstruct vision-locomotion loop from the human-prosthesis interface, this research found another way: integrating an “eye” (vision system) with intelligent prosthesis. Via this way, the prosthesis and environments could be connected, which helped amputees to reconstruct a complete vision-locomotion loop.
The designed intelligent prosthesis included a subvision system and a corresponding real-time locomotion control system, which could coordinate with the human vision system. The human vision system could perceive the environment and provided the brain with the necessary information to make decisions, but amputees could not transmit the neural signal to the prosthesis directly. The subvision system, which consisted of a depth camera, IMU, and corresponding software, could also perceive the environment actively. Then, environmental recognition results could be utilized to control the prosthesis motion. The fusion of the human vision system, the subvision system, and the locomotion control system reconstructed a complete vision-locomotion loop for an amputee. This loop assisted the amputee in achieving non-rhythmic locomotion through estimating environmental types and parameters and adjusting the control parameters accordingly.
The proposed system was validated by the experiments in which the amputees were requested to wear the powered transfemoral prosthesis to perform two tasks: 1) walking in daily terrains (level ground, up/down stairs, and up/down ramps) and 2) crossing the static and dynamic obstacles. Compared to the passive prosthesis and powered prosthesis without a vision system, the proposed intelligent powered transfemoral prosthesis had higher environmental adaptability and was able to help amputees to walk in complex environments more conveniently.
Intelligent powered transfemoral prosthesis with an “eye” helped transfemoral amputees to walk in complex environments.
The first author of this paper is Kuangen Zhang, a Ph.D. student at MEE, SUSTech. Professor Chenglong Fu is the corresponding author of this paper. This work was supported in part by the National Key R&D Program of China, the National Natural Science Foundation of China, the Guangdong Innovative and Entrepreneurial Research Team Program, the Shenzhen and Hong Kong Innovation Circle Project, and the Centers for Mechanical Engineering Research and Education at MIT and SUSTech.
The paper published on IEEE Transactions on Neural Systems and Rehabilitation Engineering proposed an unsupervised cross-subject adaptation method to predict human locomotion intent whose signals were not labeled, which liberated subjects and researchers from labeling a large amount of data.
Accurately predicting human locomotion intent is beneficial in controlling wearable robots and in assisting humans to walk smoothly on different terrains. Traditional methods for predicting human locomotion intent require collecting and labeling the human signals and training specific classifiers for each new subject, which introduces a heavy burden on both the subject and the researcher.
In addressing this issue, the present study designed an unsupervised cross-subject adaptation method to predict the locomotion intent of a target subject whose signals were not labeled. The adaptation was realized by designing two classifiers to maximize the classification discrepancy and a feature generator to align the hidden features of the source and the target subjects to minimize the classification discrepancy. A neural network was trained by the labeled training set of source subjects and the unlabeled training set of target subjects. Then it was validated and tested on the validation set and the test set of target subjects. Experimental results in the leave-one-subject-out test indicated that the present method could classify the locomotion intent and activities of target subjects at the averaged accuracy of 93.60% and 94.59% on two public datasets. The potential of the present method to predict the locomotion intent of subjects with disabilities and control the wearable robots will be evaluated in future work.
The framework of the unsupervised cross-subject adaptation for predicting human locomotion intent.
The first author of this paper is Kuangen Zhang, a Ph.D. student at MEE, SUSTech. Professor Chenglong Fu is the corresponding author of this paper. This work was supported in part by the National Key R&D Program of China, the National Natural Science Foundation of China, the Guangdong Innovative and Entrepreneurial Research Team Program, the Shenzhen and Hong Kong Innovation Circle Project, and the Centers for Mechanical Engineering Research and Education at MIT and SUSTech.
The paper published on Mechanism and Machine Theory investigated the problem of saving energy for load carriage in human walking.
Walking with heavy loads is a common task in military affairs and daily life, which can induce a significant increase in energy expenditure. Inspired by the shoulder pole for load carriage, Professor Fu’s team investigated the energy performance of human load carriage with a variable damped suspended backpack. A vertical double-mass coupled-oscillator model was adopted to predict the load movement and energy consumption. Optimal control was found to be bang-bang, switching four times between the minimum and the maximum damping in a step cycle. With analytically tested laws for optimal switching timing, an orbital stable event-based control strategy with feedback was designed. Simulation results revealed that the suspended backpack with an optimally controlled variable damper improved the energy efficiency of load carriage and reduced the load force exerting on carriers for a wide range of suspension stiffness and walking conditions, representing less energy expenditure, better comfort and stronger adaptability than the passive elastic backpacks with constant suspension parameters.
A simple model of human walking with a variable damped elastically suspended backpack.
The first author of this paper is Lianxin Yang, a Ph.D. student at ME, Tsinghua University. Professor Chenglong Fu is the corresponding author of this paper. This work was supported in part by the National Key R&D Program of China, National Natural Science Foundation of China, Guangdong Innovative and Entrepreneurial Research Team Program, Shenzhen and Hong Kong Innovation Circle Project, and Centers for Mechanical Engineering Research and Education at MIT and SUSTech.
This paper was selected as a further reading article (https://www.nature.com/articles/375052a0).
The paper published on ASME Journal of Mechanisms and Robotics designed a quasi-passive lower limb exoskeleton that was able to recycle, save, and reuse energy from knee and ankle joints to assist push-off, which provided a new way for designing an energy-efficient exoskeleton.
Humans consume more energy than the necessary energy for walking, and the redundant part has to be wasted. If the wasted energy could be recycled and released at a suitable time to push human forward, it would greatly decrease the consuming energy of the exoskeleton. The designed exoskeleton in this paper recycled the negative work performed by the knee joint in the late swing phase and the ankle joint in the mid-stance phase, to assist ankle push-off in the late-stance phase when a burst of positive power was needed. The above process was realized by using a torsion spring as an energy storage element and two clutches attached to both ends of the spring to control the timing of recycling and releasing energy in a gait cycle. The experimental results showed that the ratio of the energy released at push-off to that recycled from the knee and ankle joints was about 77.34%.
Joint power during human walking and mechanical design of the quasi-passive exoskeleton.
The co-first authors of this paper are Yihua Chang and Weixin Wang, two graduate students at ME, Tsinghua University. Professor Chenglong Fu is the corresponding author of this paper. This work was supported in part by the National Key R&D Program of China, National Natural Science Foundation of China, Guangdong Innovative and Entrepreneurial Research Team Program, Shenzhen and Hong Kong Innovation Circle Project, and Centers for Mechanical Engineering Research and Education at MIT and SUSTech.
Related paper links:
IEEE Transactions on Cybernetics：
https://ieeexplore.ieee.org/document/9042836
IEEE Transactions on Neural Systems and Rehabilitation Engineering：
https://ieeexplore.ieee.org/document/8960271
Mechanism and Machine Theory：
https://doi.org/10.1016/j.mechmachtheory.2019.103738
ASME Journal of Mechanisms and Robotics：
https://doi.org/10.1115/1.4046835
A simple model of human walking with a variable damped elastically suspended backpack.
The first author of this paper is Lianxin Yang, a Ph.D. student at ME, Tsinghua University. Professor Chenglong Fu is the corresponding author of this paper. This work was supported in part by the National Key R&D Program of China, National Natural Science Foundation of China, Guangdong Innovative and Entrepreneurial Research Team Program, Shenzhen and Hong Kong Innovation Circle Project, and Centers for Mechanical Engineering Research and Education at MIT and SUSTech.
This paper was selected as a further reading article (https://www.nature.com/articles/375052a0).
The paper published on ASME Journal of Mechanisms and Robotics designed a quasi-passive lower limb exoskeleton that was able to recycle, save, and reuse energy from knee and ankle joints to assist push-off, which provided a new way for designing an energy-efficient exoskeleton.
Humans consume more energy than the necessary energy for walking, and the redundant part has to be wasted. If the wasted energy could be recycled and released at a suitable time to push human forward, it would greatly decrease the consuming energy of the exoskeleton. The designed exoskeleton in this paper recycled the negative work performed by the knee joint in the late swing phase and the ankle joint in the mid-stance phase, to assist ankle push-off in the late-stance phase when a burst of positive power was needed. The above process was realized by using a torsion spring as an energy storage element and two clutches attached to both ends of the spring to control the timing of recycling and releasing energy in a gait cycle. The experimental results showed that the ratio of the energy released at push-off to that recycled from the knee and ankle joints was about 77.34%.
Joint power during human walking and mechanical design of the quasi-passive exoskeleton.
The co-first authors of this paper are Yihua Chang and Weixin Wang, two graduate students at ME, Tsinghua University. Professor Chenglong Fu is the corresponding author of this paper. This work was supported in part by the National Key R&D Program of China, National Natural Science Foundation of China, Guangdong Innovative and Entrepreneurial Research Team Program, Shenzhen and Hong Kong Innovation Circle Project, and Centers for Mechanical Engineering Research and Education at MIT and SUSTech.
Related paper links:
IEEE Transactions on Cybernetics：
https://ieeexplore.ieee.org/document/9042836
IEEE Transactions on Neural Systems and Rehabilitation Engineering：
https://ieeexplore.ieee.org/document/8960271
Mechanism and Machine Theory：
https://doi.org/10.1016/j.mechmachtheory.2019.103738
ASME Journal of Mechanisms and Robotics：
https://doi.org/10.1115/1.4046835