Quantitative Comparison of Grasp Qualities of Two Tendon-driven Hands Using a Novel Methodology

While there is great interest in designing and controlling biologically-inspired, tendon-driven robotic hands, we have lacked computational methodologies to evaluate and refine alternative topologies based on objective grasp metrics. We have overcome this obstacle by employing computational geometry to find the full set of feasible grasp wrenches and the corresponding grasp quality metrics for a collection of fingers driven by multiple tendons with arbitrary routing and maximal tensions. We present this analysis for twoand three-finger grasps performed by threedimensional fingers, each with 4 kinematic degrees of freedom (DOFs, one ad-abduction and three flexion-extension joints). Introduction: Biologically-inspired tendon-driven systems have been designed for the past few decades for the purposes of grasping and manipulation [1]. These systems have distinct advantages over torquedriven systems including light weight, low backlash, small size, high speed, remote actuation, and significant design flexibility in setting moment arms and maximal tendon tensions [2]. This allows optimization of system output capabilities for a particular task. Here we demonstrate a comprehensive technique for computing the full set of feasible grasp wrenches for arbitrary tendon-driven finger topology and grasp configuration, allowing the calculation of grasp quality metrics. We present comparisons of grasp quality for two different tendon-driven finger topologies and two grasp configurations. Methods: The tendon layouts analyzed are shown in Figure 1. The link lengths were 2cm and all moment arms were 5mm. Figure 1a shows a “2N” design which has 8 tendons (N is the number of DOFs) and Figure 1b shows an “N+1” design which has 5 tendons. The maximal tendon tension sum is 1000N, which is divided up evenly among the tendons for each finger. Figure 1: Tendon layouts analyzed. (a) 2N design. (b) N+1 design. The fingertip force-production capabilities for these two designs are determined by calculating the feasible force set, which is a function of tendon layout [3]. After the feasible force sets are calculated, they are intersected with friction cones in order to produce a feasible object force set. This set represents the forces that can be applied to the object by the fingertip, and it is illustrated in Figure 2. Figure 2: Feasible force set intersected with friction cone. Joint 1 Ad-abduct Joint 2 Flex-extend Joint 3 Flex-extend Joint 4 Flex-extend