Energetic and Dynamic Performance Enhancements for Compliant Robot Actuation

The vast repertoire of human skills enable nimble and graceful execution of several habitual tasks. On the contrary, despite many advances, robots are still far from matching human capabilities. Robots can be envisioned to take up the oft-touted dull, dirty and dangerous humans jobs only with breakthroughs in their energetic and dynamic characteristics. A typical bio-inspired approach that can potentially help robots achieve the aforesaid performance aspects is joint compliance. Compliant actuation technology has vastly diversified in the last decade. A survey of the various actuator methodologies hints at possible advances for series-clutched and series-parallel multi-articulated actuation. To address them, the thesis builds upon these existing actuation concepts and focuses on specifically improving their energetic and dynamic performance characteristics. Energetic benefits of series elastic actuation is often marred by gearbox friction. An instance is when friction dissipates the link energy impeding gravity-driven link motions. At such instances, clutches can help stem undesired power-flows by decoupling the link and motor. In addition, when natural link motions are to be damped while not driving the actuator, clutches can be used to actively exploit slippage to dissipate the excess mechanical energy. Such a continuous clutching action has significant implications for energy economy. Therefore, first contributions of this thesis is towards deriving an energy-based model and an optimal controller for a series clutched actuator. In addition, an optimization-based approach is sought to obtain design parameters for a physical implementation of the actuator. Parallel actuators can often augment series elastic actuators as secondary torque sources. Owing to their energy storage capabilities, they have been used to greatly enhance robot energetics. However, the joint torque resolution problem when employing dissimilar (series and parallel) actuators is difficult, more so when the parallel actuators are biarticulated. Therefore, a second contribution of the thesis is towards deriving an energetic criteria-driven, torque resolution controller. An energetic analysis of the various criteria predicts that allocation of maximum possible torque demand to the higher efficiency actuators may not necessarily be the best strategy at all times. The analysis when extended to a wider range of motion frequencies predicts progressively lesser utilization of parallel actuators can contribute to higher energy-economy. Through energy-recycling, mono and biarticulated parallel compliance can amplify jumping performance of series elastic actuated robots. While the principle augments robot performance from the mechanical domain, joint velocities powered by series actuators yet suffer from limited voltage. Field weakening is applied at the electrical level to alleviate voltage constraints. In order to maximize energetic economy during such highly dynamic motions, trajectory optimization is further employed with knowledge of actuation capabilities and novel power constraints. Therefore, the confluence of these methods is proposed and experimentally demonstrated to significantly enhance jumping performance. Finally, the efficacy of these concepts towards enhancing explosive motions are quantified through a centroidal dynamic manipulability analysis. These results lead to a third broad contribution from this thesis.