As a novel concept in the physical human-robot interactions (pHRI), the variable-stiffness-link (VSL) robot is still far from real-world applications. To narrow the gaps between the VSL robots and applications, several related topics are investigated in this dissertation. Firstly, several models describing the VSL dynamics and impact dynamics are developed. Based on such models, two design studies are presented to determine how to maximize the benefits of VSL robots in pHRI by optimizing the dynamics parameters. Secondly, a robust control design scheme is proposed to address the robust tracking and vibration suppression of VSL robots. The control design works for both single-link and multi-link VSL robots. The efficacy of the control design is validated by experimental and simulation results. Thirdly, the safe trajectory planning problem is studied. For single-link VSL robots, a strategy based on an optimal control theory is proposed. The experimental results validate that the VSL robot can have a better safety performance and a faster running speed compared with a traditional robot. For multiple-link VSL robots, human motion prediction and impact prediction are integrated in the planning framework. The proposed strategy considers the human motion in the prediction of the impact, which can improve the quality of the impact prediction. Simulation shows that the planning can avoid unnecessary robot decelerations during operations. Also, with the impact information, the planner can enforce the safety in pHRI. These topics are integrated properly in this dissertation to facilitate the applications of VSL robots in pHRI.