The force-production and passive stiffness capabilities of fingers are two critical design specifications for dexterous robotic hands. We used the link and joint kinematic parameters of the 4-DOF DLR index finger to explore the tradeoff between these two design specifications as a function of the number, routing, stiffness, and strength of each tendon. Our innovative computational approach allowed building the Pareto front of optimized passive endpoint stiffness (measured by the eccentricity of the endpoint stiffness ellipsoids) vs. maximal force-production capabilities (measured by the size and shape of the force polytope) for 1,200 randomly generated valid routings with 5, 6, 7, or 8 tendons. Our results show that this parametric optimization can increase realizable isotropic forces by up to 80% compared to the default tendon tension distribution. In addition, designs with 5 or 6 tendons can have endpoint stiffness ellipsoids with optimized low eccentricities and with force production capabilities comparable to designs with 7 or 8 tendons. Interestingly, we did not find a systematic tradeoff between force-production and passive stiffness capabilities, given a specific routing. However, the choice of number, routing and strength of each tendon greatly affects force and passive stiffness capabilities of robotic finger, which reveals the many design opportunities afforded by tendon-driven manipulators and offers insight into the anatomical features of the human musculoskeletal system.
展开▼
