One of the problems common to tendon-driven anthropomorphic robot hands is the dependency problem. Dependency arises when guiding tendons around joints not through the center of articulation which makes the length of the tendon paths for some joints depend upon the angle of other joints. Instead of the bulky mechanical solutions for this problem, this work proposes handling this problem at the software level. The core of the solution is a mapping-function that associates the desired joint angles to the correct servomotor angles accounting for all the dependencies in the system. The geometrical analysis to get this function is difficult due to the complexity of the paths of the tendons around the phalanxes of the fingers. This work proposes getting this function through learning. Using model based learning requires the equally complex analysis to build the model. Therefore, this work proposes learning by interacting with the real physical system. To evaluate the system a simple setup of an anthropomorphic robot hand was developed and used to evaluate the performance of the proposed technique. The test was done on the index finger.
M. Shaaban, K., & F. Aljunaid, A. (2007). Reinforcement Learning for Designing Tendon-Driven Anthropomorphic Robot Hand. JES. Journal of Engineering Sciences, 35(No 3), 713-728. doi: 10.21608/jesaun.2007.113994
MLA
Khaled M. Shaaban; Abdulmajeed F. Aljunaid. "Reinforcement Learning for Designing Tendon-Driven Anthropomorphic Robot Hand", JES. Journal of Engineering Sciences, 35, No 3, 2007, 713-728. doi: 10.21608/jesaun.2007.113994
HARVARD
M. Shaaban, K., F. Aljunaid, A. (2007). 'Reinforcement Learning for Designing Tendon-Driven Anthropomorphic Robot Hand', JES. Journal of Engineering Sciences, 35(No 3), pp. 713-728. doi: 10.21608/jesaun.2007.113994
VANCOUVER
M. Shaaban, K., F. Aljunaid, A. Reinforcement Learning for Designing Tendon-Driven Anthropomorphic Robot Hand. JES. Journal of Engineering Sciences, 2007; 35(No 3): 713-728. doi: 10.21608/jesaun.2007.113994
)
View on SCiNiTO