Rigid vs. elastic actuation: Requirements & performance

Intrinsically elastic joints have become increasingly popular over the last years. Commonly, they are considered to outperform rigid actuation in terms of peak dynamics, robustness, and energy efficiency. In particular, the possible increase of link speed by adequate motor excitation trajectories, such that the elastic transmission temporarily stores elastic energy and then timely converts it into kinetic link energy, is a new control problem in robotics. However, despite being a popular argument in favor of elastic actuation, it was not shown yet that this potential speed gain is truly inherent to the physical properties of the mechanism. In order to argue that 'elasticity is superior to input torque', i.e. size and weight, it still needs to be derived that this new feature does not come at the cost of increasing weight for a given actuation technology. Therefore, we analyze, under which circumstances 'extracting' a certain amount of mass from a rigid joint and 'investing' this into an elastic mechanism in the drive train leads to such a performance increase. For this, we derive the general scaling behavior of rigid joints and compare their capabilities in terms of maximum velocity to the performance behavior of an elastic joint, while taking into consideration the most important real-world constraints.