We designed, created and tested an underactuated soft gripper able to hold everyday objects of various shapes and sizes without using complex hardware or control algorithms, but rather by combining sheets of flexible plastic materials and a single servo motor. Starting with a prototype where simple actuation performs complex varied gripping operations solely through the material system and its inherent physical computation, the paper discusses how embodied computation might exist in a material aggregate by tuning and balancing its morphology and material properties.
We can define embodied computation as information processing in which the physical realization and the physical environment play an unavoidable and essential role . In this paper we will discuss embodied computation and suggest a material system that has reduced actuation complexity and performs gripping instead through an embodied material computation. Human-robot Interaction can manifest indirectly, in the sense that robots should be able to interact with the same environments humans do. This requires a certain resemblance between robots and humans: in behavior, morphology, materiality, and scale. But how do we determine what similarities relevant, and should we mimic or replicate these mechanisms? What aspects of embodied computation are relevant to the design of material systems, morphology, and material behavior? One major challenge in robotics is picking up and holding everyday objects without crushing them. For that we have created an adaptive, robust gripper able to interact with a large number of real objects from an office environment and with humans. Traditionally in computer-science, software has been developed and analyzed separately from hardware. In embodied computing the computation is seen as happening ”as a physical system in continuing interaction with other physical systems (its environment)”. . Information processing is implicit here because the physical environment performs some computations for free. Redstr¨om argues that computers can be seen as a kind of material, and that their computational capabilities must be combined with other kinds of materials in order to create a computational composite, so that the computer becomes useful in design . This paper contributes a design of an underactuated gripper, a computational aggregate made up of material composites in a soft mechanical system, with an emphasis on morphology and material behavior interacting with the real environment. II. RELATED WORK Many robotic hand designs focus on mechanically replicating the human hand, controlling each joint independently using many actuators. On the other hand, underaction designs employ less actuators in order to control a larger number of joints. One of the first examples of an underactuated soft gripper, similar to a bicycle chain, was developed using pulleys and twenty articulations. It was able to conform objects of arbitrary shapes . However, the design of this gripper only permitted holding an object in one plane. Another example of a soft universal gripper could conform around a complex object from all sides, and hold it by contracting the granular material it was made from . This is a good example of embodied computing where a computational composite is used. The granular material automatically computes and shapes around an object, simplifying and avoiding the problems multifingered robotic hands experience when needing to compute the force and position required to control each finger. A simple design employing material intelligence can thus avoid both hardware and software complexities.