Maximizing energy efficiency of variable stiffness actuators through an interval-based optimization framework

Abstract

Despite the necessity for being able to regulate the stiffness of the robotic platforms in physical contact with humans, and tremendous number of different variable stiffness actuators that have been developed so far, there is still no such actuator that has successfully passed the research lab phase and transferred into a real application. The main reason is due to the lack of understating on how to optimally design a stiffness adjustment mechanism based on desired performances of each application. If not optimally designed, the additional complexities within the actuators, prevent to win the trade-off between benefits of having a stiffness adjustment mechanisms versus its costs, such as the energy storage capacity of the elastic elements compared to how much energy can be actually released at the output of the actuator, i.e. the link. Currently, this trade-off criterion is not in favor of introducing a stiffness adjustment mechanism into the actuator. Therefor, generally, having a simple series elastic actuator with active compliance has been preferred over having a complex variable stiffness actuator, in many real applications. This work develops an understanding of how to optimally design the parameters of a stiffness adjustment mechanism by developing a framework that can robustly maximize the energy efficiency of variable stiffness actuators. Five different design sets of stiffness adjustment mechanism are being considered and evaluated based on the proposed optimization framework. The resultant optimal design of each set is then compared with the original design in terms of energy efficiency. The proposed framework shows improvement of energy efficiency up to 354%, while design constraints are all being satisfied.