A Variable Stiffness Actuator (VSA) is an inherently parameter-dependent system due to the controllable stiffness element. Within the torque-controlled framework, the VSA is distinguished from the classical Series Elastic Actuator (SEA). A frozen torque controller can directly determine the SEA performance with a fixed-stiffness spring selection. However, since the VSA operates at a set of operating points, the aim is to achieve an adaptive control approach. For this, we propose a gain-scheduled torque controller. The control performance is expected to recover robustness when stiffness values are varied from smaller to larger ones. Simultaneously the bandwidth is maximized, taking into account hardware limitations. In this way, a good trade-off between stability and performance can be achieved. A key step in the gain-scheduled controller design is to implement the linear controllers, where the Linear Quadratic Gaussian (LQG) technique was applied to deal with the multiloop feedback (a cascade control scheme) in the VSA plant. A lever-arm based VSA (i.e. a mechanical-rotary variable impedance actuator (MeRIA)) was used to verify this new controller design approach by the simulations and experiments. The resulting gain-scheduled controller was also evaluated by taking into account the impedance control testing on a motion-supported platform for the knee joint.

This paper investigates a closed-loop torque-controlled variable stiffness actuator (VSA) combined with a disturbance observer for enhancing low output impedance. We implement the generalized extended state observer (GESO) for conveniently testing the stability of the time-varying VSA system. In our application, the GESO is also required to serve the operation of the low and high impedance task. Here, the most important aspect is to consider the influence of the physical stiffness on the output impedance, because the VSA has been regulated with the closed-loop torque control. Through the interaction-torque experiments, we verify that using the fast dynamics GESO with the low-stiffness actuator can achieve low output impedance and stable interaction under the reachable frequency of a human. These properties contribute to perform the low-impedance task. When performing the high-impedance task, where a large torque command is needed, the high-stiffness actuator and the slow dynamics GESO are implemented to achieve high bandwidth and proper tracking performance. The continuously variable observer responses in accordance with the stiffness values are achieved via the gain-scheduling control. Moreover, the present closed-loop linear parameter varying system can be verified to be quadratically stable. The VSA system is then implemented on a knee exoskeleton for a sit-to-stand application. The reference command of the exoskeleton is a joint torque, calculated from the inverse dynamics. This torque signal is also used as the reference command of the stiffness motor of the VSA. The effectiveness of the exoskeleton system is experimentally verified with one healthy volunteer. Subsequently, another two healthy volunteers also successfully experienced the system.

We present a framework for achieving a robust stability test of the variable stiffness actuator (VSA) exoskeleton programmed with the gain scheduling-based variable impedance control (GSVIC). In this brief, the focused impedance control framework involves the cascaded position torque control loop, in which both control loops are closed, and the index of the GSVIC is related to the joint torque of the human. However, there is a lack of research on the variation rate of such a biofeedback signal. This lack of information is necessary when performing the robust stability test of the linear parameter-varying (LPV) system. To acquire a bounded variation rate, the mechanical stiffness variation component of the VSA is used to transfer the biofeedback information into the GSVIC. In this case, the LPV system has only one variable parameter with the known and bounded variation rate, i.e., the physical stiffness of the VSA. The operating range of the impedance controller, determined from the test of robust stability via parameter-dependent Lyapunov functions, can meet our application requirements. The whole control system follows the paradigm of the variable impedance task in accordance with human intention. The effectiveness of the control scheme is experimentally verified on a VSA prototype and two subjects wearing the VSA exoskeleton.

Extracting cardiorespiratory signals from non-invasive and non-contacting sensor arrangements, i.e. magnetic induction sensors, is a challenging task. The respiratory and cardiac signals are mixed on top of a large and time-varying offset and are likely to be disturbed by measurement noise. Basic filtering techniques fail to extract relevant information for monitoring purposes. We present a real-time filtering system based on an adaptive Kalman filter approach that separates signal offsets, respiratory and heart signals from three different sensor channels. It continuously estimates respiration and heart rates, which are fed back into the system model to enhance performance. Sensor and system noise covariance matrices are automatically adapted to the aimed application, thus improving the signal separation capabilities. We apply the filtering to two different subjects with different heart rates and sensor properties and compare the results to the non-adaptive version of the same Kalman filter. Also, the performance, depending on the initialization of the filters, is analyzed using three different configurations ranging from best to worst case. Extracted data are compared with reference heart rates derived from a standard pulse-photoplethysmographic sensor and respiration rates from a flowmeter. In the worst case for one of the subjects the adaptive filter obtains mean errors (standard deviations) of -0.2 min ⁻¹ (0.3 min ⁻¹) and -0.7 bpm (1.7 bpm) (compared to -0.2 min ⁻¹ (0.4 min ⁻¹) and 42.0 bpm (6.1 bpm) for the non-adaptive filter) for respiration and heart rate, respectively. In bad conditions the heart rate is only correctly measurable when the Kalman matrices are adapted to the target sensor signals. Also, the reduced mean error between the extracted offset and the raw sensor signal shows that adapting the Kalman filter continuously improves the ability to separate the desired signals from the raw sensor data. The average total computational time needed for the Kalman filters is under 25% of the total signal length rendering it possible to perform the filtering in real-time. It is possible to measure in real-time heart and breathing rates using an adaptive Kalman filter approach. Adapting the Kalman filter matrices improves the estimation results and makes the filter universally deployable when measuring cardiorespiratory signals.

The cascade impedance control (CIC) is a wellknown architecture framework for rehabilitation exoskeleton with compliant joint (e.g., a series elastic actuator). The cascaded position-torque-velocity control loop of the CIC system can deal with specific issues, such as stiction or the accuracy of output impedance. In this paper, the control loops of the CIC are extended by the traditional iterative learning control (ILC), and then we examine and compare three types of frameworks, named torque learning, impedance learning, and trajectory learning. Their advantages, such as reducing the lag of the output trajectory with a low-gain impedance controller (safety), are discovered. Furthermore, the exoskeleton system is upgraded with the ability of a variable impedance. In this part, a fuzzy logic system is proposed. This system employs the electromyography signals of the subject and the exoskeleton torque as input, and the decisions on the variable impedance as output. The experiment verifies that the proposed algorithm can effectively decrease the impedance of the exoskeleton when detecting a spasticity from the subject, and can maintain the original dynamics of the system when the subject performs a normal movement. Afterwards, the effectiveness of the fuzzy logic system together with the ILC-based CIC is experimentally verified in the case of subject-exoskeleton collaborative.