Conceptualization of an exoskeleton Continuous Passive Motion(CPM) device using a link structure

Intention Detection Using Physical Sensors and Electromyogram for a Single Leg Knee Exoskeleton

In this paper, we present a knee exoskeleton. Due to the complicated structure of the knee, an exoskeleton can limit the wearer’s movement (e.g., when completely sitting down). To prevent this, the proposed exoskeleton is designed to move the ankle part prismatically, so the movement of the wearer is not limited. In addition, the developed exoskeleton could be worn on only one leg, but in this case, it is difficult to detect the intention because the relative relationship information of the two legs is unknown. For this purpose, the length between the knee center of rotation and the ankle (LBKA) was measured and used for intention detection. Using a physical sensor—an encoder and an LBKA sensor, the success rate of intention detection was 82.1%. By additionally using an electromyogram (EMG) sensor, the success rate of intention detection was increased to 92%, and the intention detection was also 27.1 ms faster on average.

Development of a Single Leg Knee Exoskeleton and Sensing Knee Center of Rotation Change for Intention Detection

In this study, we developed a single leg knee joint assistance robot. Commonly used exoskeletons have a left-right pair, but when only one leg of the wearer is uncomfortable, it is effective to wear the exoskeleton on only the uncomfortable leg. The designed exoskeleton uses a lightweight material and uses a wire-driven actuator, which reduces the weight of the driving section that is attached on the knee directly. Therefore, proposed exoskeleton reduces the force of inertia that the wearer experiences. In addition, the lower frame length of the exoskeleton can be changed to align with the complex movement of the knee. Furthermore, the length between the knee center of rotation and the ankle (LBKA) is measured by using this structure, and the LBKA values are used as the data for intention detection. These value helps to detect the intention because it changes faster than a motor encoder value. A neural network was trained using the motor encoder values, and LBKA values. Neural network detects the intention of three motions (stair ascending, stair descending, and walking), Training results showed that intention detection was good in various environments.