Seri tipteki değişken esnekliğe sahip eyleyicilerin eş-tasarım eniyileştirmesi

Variable stiffness actuation (VSA) improves the dexterity of robots while ensuring the safety of physical human-robot interactions. VSAs can be implemented based on several principles. Among these, lever-type VSAs are advantageous, since no mechanical work against the spring is required to modulate the stiffness of the system under ideal conditions. We study co-design optimization of lever-type VSAs where mechanical and control design variables are simultaneously optimized to achieve the best performance. We demonstrate that co-design optimization results in significant performance advantages in comparison to subsystem optimization. Given that VSA optimization problems inherently consider multiple criteria, we model the co-optimization problem accordingly and compute all non-dominated system-level optima as a Pareto-front curve. We provide sample design selections from the Pareto-front curve and discuss the trade-offs involved in various solutions. We also provide two single-actuator solutions: a series elastic actuation solution and a novel constant-torque solution that achieves the task through stiffness regulation. Through a modified optimization formulation, we further demonstrate that the constant torque requirement for the constant torque solution can be decreased at the expense of the lower stiffness of the actuator. Finally, we compare Pareto solutions of lever-type VSA mechanisms with those of single spring preload adjustment based VSA, for the same task and objective functions. Through this comparison, we provide evidence that series type VSA or single spring preload based VSA can be favorable for different designer preferences.