Probabilistic Elastic Element Design for Robust Natural Dynamics of Structure-Controlled Variable Stiffness Actuators

This paper proposes a probabilistic approach for the design of elastic elements to be used in structure-controlled variable stiffness actuators (VSA) for robotic applications. Considering the natural dynamics of the elastic actuation system, requirements are defined and material selection as well as geometry calculation are performed using lumped parameter models. Monte Carlo simulations are integrated in the design procedure to ensure a robust implementation of the required dynamical characteristics. Thereby, effects of uncertainties that might be caused by manufacturing or deviations of material properties are taken into account. To validate the suitability of the overall approach and the particular methods, a torsional elastic element is implemented and experimentally evaluated. The evaluation shows a fulfillment of the key requirements, i.e., specific natural dynamic behavior, that is only achieved due to considering uncertainties. Further, the transferability of the approach to other structure-controlled elastic actuators is discussed and implications are given. Only the governing equations of stiffness properties in certain load situation need to be adapted, e.g., from torsion to bending. Due to the simple transfer, the proposed probabilistic and model-based approach is promising for application to various actuator concepts with structure-controlled variable stiffness.