In this paper, the design and experimental validation of a knee exoskeleton are presented. The exoskeleton can capture the negative work from the wearer’s knee motion while decreasing the muscle activities of the wearer. First, the human knee biomechanics during the normal walking is described. Then, the design of the exoskeleton is presented. The exoskeleton mainly includes a left one-way transmission mechanism, a right one-way transmission mechanism, and a front transmission mechanism. The left and right one-way transmission mechanisms are designed to capture the negative work from the wearer’s knee motion in the stance and swing phases, respectively. The front transmission mechanism is designed to transform the bidirectional rotation of the wearer’s knee joint into the generator unidirectional rotation. Additionally, the modeling and analysis of the energy harvesting of the exoskeleton is described. Finally, walking experiments are performed to validate the effectiveness of the proposed knee exoskeleton. The testing results verify that the developed knee exoskeleton can output a maximum power of 5.68 ± 0.23 W and an average power of 1.45 ± 0.13 W at a speed of 4.5 km/h in a gait cycle. The average rectus femoris and semitendinosus activities of the wearers in a gait cycle are decreased by 3.68% and 3.40%, respectively.

Robotic assistive or rehabilitative devices are promising aids for people with neurological disorders as they help regain normative functions for both upper and lower limbs. However, it remains challenging to accurately estimate human intent or residual efforts non-invasively when using these robotic devices. In this article, we propose a deep learning approach that uses a brightness mode, that is, B-mode, of ultrasound (US) imaging from skeletal muscles to predict the ankle joint net plantarflexion moment while walking. The designed structure of customized deep convolutional neural networks (CNNs) guarantees the convergence and robustness of the deep learning approach. We investigated the influence of the US imaging’s region of interest (ROI) on the net plantarflexion moment prediction performance. We also compared the CNN-based moment prediction performance utilizing B-mode US and sEMG spectrum imaging with the same ROI size. Experimental results from eight young participants walking on a treadmill at multiple speeds verified an improved accuracy by using the proposed US imaging + deep learning approach for net joint moment prediction. With the same CNN structure, compared to the prediction performance by using sEMG spectrum imaging, US imaging significantly reduced the normalized prediction root mean square error by 37.55% ($ p $ < .001) and increased the prediction coefficient of determination by 20.13% ($ p $ < .001). The findings show that the US imaging + deep learning approach personalizes the assessment of human joint voluntary effort, which can be incorporated with assistive or rehabilitative devices to improve clinical performance based on the assist-as-needed control strategy.

Musculoskeletal models, like all theoretical models of physical processes, depend on the assumptions needed to construct the model. For musculoskeletal models, these assumptions include, among other things, the kinematic data, the kinetic data and the muscle parameters. The former (dynamic) data can be acquired relatively easily from living subjects, but the latter are usually based on limited information, frequently determined from cadaver studies performed on elderly individuals. Previously, we determined the sensitivity of forces to dynamic differences among 10 humans walking on a straight path. Here, we assess the sensitivity of the muscle and joint reaction forces developed in human walking to variable muscle parameters obtained from 10 living adults, whose data were recently reported, and compared the results with the values from a standard model that depends on cadaveric data. We found that, while the force patterns across the stance cycle were similar among muscle parameter models, differences of as much as 15% in the force magnitude were produced. Whether or not the variation between the standard model and other muscle parameters is important depends on why the forces are required.