The aim of this workshop is to help in the growing need of mathematical tools and shared standards in the emergent field of soft manipulation. In particular, it will tackle the problem of finding innovative ways to evaluate grasps with underactuated and soft robotic hands. The development of such devices is a cutting-edge research topic that requires an interdisciplinary approach. In this context, having a common benchmarking framework for assessing the quality of compliant and underactuated manipulation systems could foster fruitful scientific collaborations, drive the design of new robotic hands and facilitate the grasp execution process with existing hands.
Benchmarking robotic hands is particularly problematic due to the complexity of these mechanisms and the wide range of tasks they are expected to accomplish. They can be evaluated based on their physical characteristics (weight, number of DOFs, power consumption, etc.) and on their functional performance. The problem of computing the quality of grasps performed by underactuated and compliant hands has been rarely addressed in the literature. Classical quality indexes are mainly thought for precision grasps with fully actuated hands, and either compute the optimal location of contact points on the object, or the optimal hand configuration. These concepts, however, may not be relevant when dealing with underactuated and compliant hands that adapt to the objects’ shape and are likely to perform power grasps, for which the quality evaluation based on traditional approaches has high computational costs. In addition, approximating the real contact points/areas becomes difficult when dealing with soft continuous robotic hands like the RBO Hand 2.
More recently, the use of object databases or grasp taxonomies has tried to cover the need for standardized benchmarking tasks. This workshop will not only open a debate on the application of these methods to underactuated and soft hands, but will also address how and to what extent human grasping can be used as a reference for this type of hands. The introduction of soft hands that exploit environmental constraints to perform manipulation and grasping tasks opens also new and interesting scenarios for learning from human strategies and adapting them for robotic manipulation. Also, the use of underactuation, although limiting the ability for in-hand manipulation, still offers a manipulation workspace that can be effectively exploited for different applications.
This workshop will bring together researchers with different backgrounds to exchange ideas on the topics outlined above, with inspiring talks to sparkle discussions among the assistants, contributions of posters and demos to illustrate the latest achievements in the field both from the hardware and software perspective, and with a round table and continuous debate to try to converge towards new ideas that can guide the design, planning, and control of soft manipulation.