In this paper, we propose a novel rotational variable stiffness actuator that uses S-shaped springs as elastic elements. The S-shaped springs are designed based on the principle of variable beam length, which allows for an easily customizable and wide-ranging stiffness regulation. The stiffness can be smoothly regulated online by rotating the S-shaped springs, resulting in a compact stiffness regulation mechanism. Moreover, a planetary gear differential is used to drive the stiffness regulation mechanism such that both the stiffness control motor and the main drive motor can be placed at the proximal part. This configuration effectively reduces the reflected inertia of the moving parts, which helps to achieve a compact design and improve the control performance. The proposed actuator ranks high among designs of the similar output power in terms of the power-to-weight ratio and the power-to-volume ratio; therefore, the compactness of the proposed design is verified. Besides, the stiffness regulation speed of the proposed design also ranks high among the designs. Consequently, the actuator can be used for fast stiffness regulation. For robust position tracking under different operating conditions, a disturbance observer is used to estimate the mismatched lumped disturbance caused by the load torque and model uncertainties. To further improve the control performance, a sliding mode control term is introduced to compensate for the disturbance estimation error. The stability of the closed-loop system is proved using the Lyapunov method. Experimental results under different operating conditions have verified the validity of the novel actuator and the effectiveness of the proposed controller.

An adaptive robot control approach that can achieve an accurate approximation of the complex robot dynamics without knowledge unlearning in the case of multiple sequential tasks is proposed in this article. A locally weighted learning (LWL) model with automatic structure growth according to the size of the learning domain is introduced to approximate unknown robot dynamics, and a composite learning technique with regressor extension is applied to improve parameter convergence. The LWL ensures that learning in one area of the learning domain does not cause unlearning in another area, and the composite learning theoretically guarantees the identification performance of the LWL model. By the exploitation of both stored data in memory and instantaneous data, parameter convergence in the LWL model is achieved under a more achievable interval-excitation condition than the stringent persistent-excitation condition. This further enhances the trajectory tracking performance for multiple sequential tasks, which is generally not achievable by existing approximation-based adaptive robot control approaches. Experimental studies have verified the superiority of the proposed approach over prevalent approaches.