Adaptive Intention-Driven Variable Impedance Control for Wearable Robots With Compliant Actuators

Huo, Yingxin; Li, Xiang; Zhang, Xuan; Li, Xiu; Sun, Dong

doi:10.1109/tcst.2022.3222728

Cited by 6 publications

(2 citation statements)

References 46 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Furthermore, predicting and controlling different human motion intentions innately is complex since the results are based on data from predetermined experimental motions. To address these limitations, future studies could incorporate EMG or insole sensors, as previous research has suggested [6,20,23,33], to predict motion intention in addition to the IMU sensor.…”

Section: Discussionmentioning

confidence: 99%

“…To address these issues, Akbari, M. et al [11] proposed a method to evaluate the uncertainty of a deep learning algorithm in real time and use this uncertainty measurement in the robot's control loop to update the system whenever an unsafe situation for the user occurs. In [33], an adaptive intention-based variable impedance controller was proposed to estimate human motion intentions online based on physical interactions, stochastic distributions, and random disturbances by Huo, Y. et al As such, since exoskeleton robots are worn by humans, it will be essential to solve the problem of false detections caused by algorithms. In the future, after developing a multi-modal sensor-based motion intention prediction algorithm, we will secure safety by establishing a response strategy in case of false detection of the algorithm, as in previous studies.…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

Machine Learning Based Recognition of Elements in Lower-Limb Movement Sequence for Proactive Control of Exoskeletons to Assist Lifting

Woo,

Sahithi,

Kim

et al. 2023

IEEE Access

View full text Add to dashboard Cite

Exoskeleton robots that provide physical assistance to workers at industrial sites have recently been advanced by utilizing various bio-signal measurement sensors and artificial intelligence. However, their commercialization speed is slow compared to the pace of technological development. One of the reasons for this is the motion mismatch that occurs between the control of the exoskeleton and the worker's response. To solve this issue, the worker's intended motion could be predicted, and proactive control of the exoskeleton robot could be implemented. An experiment was conducted with 35 subjects, with the data of 30 subjects used for unsupervised learning and the data of the remaining 5 subjects used for supervised learning. To predict the intended motion of the subjects, the data from IMU sensors were used to segment the motion elements through a k-means clustering algorithm, employing a motion segmentation technique. The lower extremity motion was understood as a sequence composed of motion elements. The potential for predicting motion intention and sequence was demonstrated by comparing the results from unsupervised learning with those of an MLP model that predicted motion sequences from new, unused data. Additionally, it was confirmed that proactive control of the exoskeleton robot using the motion segmentation technique was possible when the duration of the element motion constituting the lower extremity motion sequence exceeded 200ms. Therefore, this study conducted preliminary research to develop a system for predicting a worker's intended motion and motion sequence for proactive control of an exoskeleton robot.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%