AUTHOR=Haas Mark , Friedl Werner , Stillfried Georg , Höppner Hannes TITLE=Human-Robotic Variable-Stiffness Grasps of Small-Fruit Containers Are Successful Even Under Severely Impaired Sensory Feedback JOURNAL=Frontiers in Neurorobotics VOLUME=12 YEAR=2018 URL=https://www.frontiersin.org/journals/neurorobotics/articles/10.3389/fnbot.2018.00070 DOI=10.3389/fnbot.2018.00070 ISSN=1662-5218 ABSTRACT=

Application areas of robotic grasping extend to delicate objects like groceries. The intrinsic elasticity offered by variable-stiffness actuators (VSA) appears to be promising in terms of being able to adapt to the object shape, to withstand collisions with the environment during the grasp acquisition, and to resist the weight applied to the fingers by a lifted object during the actual grasp. It is hypothesized that these properties are particularly useful in the absence of high-quality sensory feedback, which would otherwise be able to guide the shape adaptation and collision avoidance, and that in this case, VSA hands perform better than hands with fixed stiffness. This hypothesis is tested in an experiment where small-fruit containers are picked and placed using a newly developed variable-stiffness robotic hand. The grasp performance is measured under different sensory feedback conditions: full or impaired visual feedback, full or impaired force feedback. The hand is switched between a variable-stiffness mode and two fixed-stiffness modes. Strategies for modulating the stiffness and exploiting environmental constraints are observed from human operators that control the robotic hand. The results show consistently successful grasps under all stiffness and feedback conditions. However, the performance is affected by the amount of available visual feedback. Different stiffness modes turn out to be beneficial in different feedback conditions and with respect to different performance criteria, but a general advantage of VSA over fixed stiffness cannot be shown for the present task. Guidance of the fingers along cracks and gaps is observed, which may inspire the programming of autonomously grasping robots.