Design of vast DOF artificial muscle actuators with a cellular array structure and its application to a five-fingered robotic hand

This paper presents a cellular actuator design for a robotic hand. Although the motions of robotic hands are complex, a design of cellular actuator segmentation can be simplified by extracting the features of given postures and using them to design and control the actuator. A method of using captured hand posture data to design the actuator segmentation is proposed. Data from the eight most frequently used hand grips in a daily living, as defined in the sollerman hand function test, is used in this design. The gathered joint angle data is transformed into actuator displacement data and used to generate a segmentation design of the actuator. For segmentation design, feature extraction method, called non-negative matrix factorization with constraints is proposed. The actuator system has 12 SMA actuators each with eight segments, resulting in 96 configurable segments before the coupled segmentation design is applied. The coupling segmentation design effectively groups the segments into clusters which are simultaneously controlled. The segmentation design reduces 96 separately controlled segments to 8, while maintaining the ability to accomplish all desired postures. A robotic hand with five fingers, designed and fabricated using the FDM process, is driven with this actuator system, and eight hand postures are reproduced with the robotic hand