In robot-assisted rehabilitation and leg exoskeletons, humans and robots are required to collaboratively complete repetitive tasks with fixed periodic paths. In such applications, impedance learning control can provide variable impedance regulation for improving the performance of physical interactions; however, designing such control is highly challenging owing to the difficulty in modeling human time-varying dynamics. By exploiting the spatial periodicity characteristics of the desired trajectory and human impedance, we propose a novel spatial repetitive impedance learning control strategy to enhance interaction performance. First, a defined spatial operator serves as the mathematics foundation for constructing the robot dynamics in the spatial domain. Then, a spatial impedance learning controller is designed. In this article, time-varying impedance profiles are estimated using spatial full-saturation iterative learning laws, while robotic parameter uncertainties are estimated using the differential adaptation law with projection modification. We validate the uniform convergence of the tracking error through a Lyapunov-like analysis and demonstrate the control effectiveness using an illustrative example. Compared with related results on temporal repetitive learning control, the proposed control approach can enable human–robot system to complete a repetitive task with unspecified speeds according to the users' strengths and motion capacity.

Model-based impedance learning control can provide variable impedance regulation for robots through online impedance learning without interaction force sensing. However, the existing related results only guarantee the closed-loop control systems to be uniformly ultimately bounded (UUB) and require the human impedance profiles being periodic, iteration-dependent, or slowly varying. In this article, a repetitive impedance learning control approach is proposed for physical human–robot interaction (PHRI) in repetitive tasks. The proposed control is composed of a proportional-differential (PD) control term, an adaptive control term, and a repetitive impedance learning term. Differential adaptation with projection modification is designed for estimating robotic parameters uncertainties in the time domain, while fully saturated repetitive learning is proposed for estimating time-varying human impedance uncertainties in the iterative domain. Uniform convergence of tracking errors is guaranteed by the PD control and the use of projection and full saturation in the uncertainties estimation and is theoretically proved based on a Lyapunov-like analysis. In impedance profiles, the stiffness and damping are composed of an iteration-independent term and an iteration-dependent disturbance, which are estimated by repetitive learning and compressed by the PD control, respectively. Therefore, the developed approach can be applied to the PHRI where iteration-dependent disturbances exist in the stiffness and damping. The control effectiveness and advantages are validated by simulations on a parallel robot in a repetitive following task.

Exoskeletons can assist the daily life activities of the elderly with weakened muscle strength, but traditional rigid exoskeletons bring parasitic torque to the human joints and easily disturbs the natural movement of the wearer’s upper limbs. Flexible exoskeletons have more natural human-machine interaction, lower weight and cost, and have great application potential. Applying assist force according to the patient’s needs can give full play to the wearer’s remaining muscle strength, which is more conducive to muscle strength training and motor function recovery. In this paper, a design scheme of an elbow exoskeleton driven by flexible antagonistic cable actuators is proposed. The cable actuator is driven by a nonlinear series elastic mechanism, in which the elastic elements simulate the passive elastic properties of human skeletal muscle. Based on an improved elbow musculoskeletal model, the assist torque of exoskeleton is predicted. An assist-as-needed (AAN) control algorithm is proposed for the exoskeleton and experiments are carried out. The experimental results on the experimental platform show that the root mean square error between the predicted assist torque and the actual assist torque is 0.00226 Nm. The wearing experimental results also show that the AAN control method designed in this paper can reduce the activation of biceps brachii effectively when the exoskeleton assist level increases.

The exoskeleton has emerged as a promising technology to enhance humans’ strength and boost users’ efficiency. In order to provide congruent human–machine interaction for assisting the users, exoskeletons should have knowledge of human’s planned action and accordingly command the robot by the designed controller. It means the ubiquitous aspect of exoskeletons lies on motion intent understanding and active compliance control. In the last decade, extensive research has been conducted on the two topics. However, no major breakthrough has been made. Thus, a systematic review and analysis on this very subject is of great significance in developing exoskeletons. Within this context, this review first surveys the history of lower limb exoskeletons to summarize the various technologies for realizing transparent human exoskeleton coordination. Then, an overview about motion intent understanding and compliance control strategies are presented in detail. Furthermore, the future trend and research directions are also outlined.

This paper proposes a Bidirectional Mamba-based selective state space model (SSM) for continuously predicting human motion intention using multisource information fusion of surface electromyogram (sEMG) and mechanomyography (MMG). In the proposed prediction model, the bidirectional canning mechanism and the dynamic selection mechanism make the prediction model have global context modeling capability and optimal allocation of computing resources, and the SSM makes the prediction model maintain linear computational complexity. By integrating the complementary features of sEMG and MMG through multimodal signal fusion, the deep information of motion intention is deeply explored, significantly enhancing the model’s intention prediction ability. The effectiveness and superiority of the proposed Bidirectional Mamba-based human motion prediction model are validated in comparison with the information fusion models based on recurrent neural network (RNN), long short-term memory network (LSTM), and temporal convolutional network (TCN).

The Schroth exercise can train the paraspinal muscles of patients with adolescent idiopathic scoliosis (AIS), however, muscle performance during the training remains unknown. This study applied surface electromyography (sEMG) to investigate the paraspinal muscle activities before, during and after Schroth exercise in nine AIS patients. This study found that after the Schroth exercise, the paraspinal muscle symmetry index (PMSI) was significantly reduced (PMSI = 1.3), while symmetry exercise significantly lowered the PMSI (PMSI = 0.93 and 0.75), and asymmetric exercise significantly increased the PMSI (PMSI = 2.56 and 1.52) compared to relax standing (PMSI = 1.36) in participants (p < 0.05). Among the four exercises, the PMSI of on all fours (exercise 1) and kneeling on one side (exercise 3) was the most and the least close to 1, respectively. The highest root mean square (RMS) of sEMG at the concave and convex side was observed in squatting on the bar (exercise 2) and sitting with side bending (exercise 4), respectively. This study observed that the asymmetric and symmetric exercise induced more sEMG activity on the convex and concave side, respectively, and weight bearing exercise activated more paraspinal muscle contractions on both sides of the scoliotic curve in the included AIS patients. A larger patient sample size needs to be investigated in the future to validate the current observations.