Search

Most control methods deployed in lower extremity rehabilitation robots cannot automatically adjust to different gait cycle stages and different rehabilitation training modes for different impairment subjects. This article presents a continuous seamless assist-as-needed control method based on sliding mode adaptive control. A forgetting factor is introduced, and a small trajectory deviation from reference normal gait trajectory is used to learn the rehabilitation level of a human subject in real time. The assistance torque needed to complete the reference normal gait trajectory is learned through radial basis function neural networks, so that the rehabilitation robot can adaptively provide the assistance torque according to subject’s needs. The performance and efficiency of this adaptive seamless assist-as-needed control scheme are tested and validated by 12 volunteers on a rehabilitation robot prototype. The results show that the proposed control scheme could adaptively reduce the robotic assistance according to subject’s rehabilitation level, and the robotic assistance torque depends on the forgetting factor and the active participation level of subjects.

Artificial intelligence-based rehabilitation robots can be applied for peoples with lower limb motor dysfunction usually caused by accident, war, sports, spinal cord injury, paralysis, and vascular diseases to enhance the motion ability of their lower limbs. A design of lower limb rehabilitation robot as well as its trajectory tracking which is suitable for passive training of patient with paralyzed or weak limbs have been presented in this paper. The contents were well organized and presented. Also, the experimental procedural setup of the lower limb rehabilitation robot is well described and conducted. Initially, AC servo motor was chosen as the actuator for controlling the robot due to good stability, fast response, low noise, wide speed range, low cost, high efficiency, and low maintenance. The corresponding simulation analysis of the actuator shows that it is good and can meet the design requirements. Secondly, the motion control of the robot was investigated using the kinematics model of the hip and knee transmission subsystem, the human hip and knee trajectory curves, the motion control data generation process, the planning target track, and the required servo motor drive pulse control instruction. Finally, the control experiments were conducted such that the pulse signal of the servo shaft rotational angle is received by the servo drive for pitch control in position control mode. By default, a complete rotation of 0.072 degrees corresponds to 500 pulses of the motor rotation time control. The experimental results show that by using PID controller, the actual trajectory is closer to the desired trajectory closely follows the actual trajectory and hence, demonstrated a promising patient's recovery.

For the problem that patients with lower limb motor dysfunction cannot generate strong exercise intention due to pain or other reasons during the rehabilitation exercise, which makes it difficult for precise control the lower limb motion assisted robot according to movement intention of patients. This paper analyzed the mechanism of virtual induction which can promote the generation of motion intention, established the virtual induction mechanism model, proposed a method to induce subjects to generate movement intention based on virtual reality (VR) technology and a lower extremity motion assisted exoskeleton robot system. The method established immersive virtual scenes that inspire the interest of subjects to induce and enhance their active movement intentions, produced electroencephalogram (EEG) signals with distinct characteristics. The experiment platform is established to perform experimental verification, and the EEG signals of subjects are collected as they move. The experimental results show that the method can effectively enhance active movement intention of subjects, it produced a stronger event-related desynchronization (ERD) phenomenon than without the virtual induction system, and the EEG signals generated have obvious characteristics after virtual induction. It lays the foundation for the movement intention perception of lower limb motion assisted robot and robot assistance control as subjects need.

Exoskeleton robots and brain computer interface (BCI) have a considerable development. But, there is a problem of relatively lower detection accuracy for voluntary movement intention. Deep learning is an effective method to solve it, hence a RP-based voluntary movement intention detection of lower limb using convolution neural network (CNN) is proposed in this paper. Firstly, the mechanism of readiness potential (RP) was analyzed, its property is an important basis for movement intention detection. Then, the lower limb voluntary movement intention detection system based on electroencephalogram (EEG) was established and six healthy subjects participated in this experiment. The different CNN models were established as a classifier. The experimental results show that the window length of EEG classification data was set as 300ms to get a better detection accuracy and time performance. The average detection accuracy of right leg voluntary movement using AlexNet, VGGNet-16 and VGGNet-19 was 72.1%, 83.0% and 84.2% respectively. Specifically, the results suggested that the detection accuracy was also increased with the increase of CNN network layer. This work demonstrates that VGGNet-based CNN model can detect lower limb voluntary movement intention, it lay a foundation for human-computer interaction of lower limb exoskeleton robot.

Real-time application of exoskeleton remains a challenge due to the limited stability of electromyogram (EMG) collected on lower limb. To enhance generalization of EMG based pattern recognition (PR), this work proposed a novel fault-tolerant algorithm based on conventional liner discriminant analysis (LDA). Based on two patterns of EMG collection (static data and dynamic data), the most accurate feature set was first selected in static data to guarantee basic performance of LDA. Detections of LDA formed a decision stream, and provided later analysis with information of movement and muscle activity of lower limb. A BP classifier was trained based on the decision stream and features set to recover misclassification of LDA when interference occurred. Both LDA and BP were trained on static data. Dynamic data were set to imitate instability of EMG in realistic application such as different force intensity and muscle fatigue. The proposed algorithm was tested on a EMG based PR system. 4 able-bodied subjects participated in the experiment. The enhancement in classification accuracy ranged from 7.52% to 13.94% over subjects compared with unoptimized system. After EMG channels reduced, this enhancement ranged from 9.01% to 27.35%, which significantly improved classification performance under limited condition. These results suggested that the proposed algorithm yield excellent performance in improving the generalization of EMG based PR.