Robot Thumb Kinematic Model Optimisation Spring Term 2012

This reports documents work undertaken in optimising a kinematic model of the human thumb. It is motivated by the high rejection rate observed with myoelectric hand prostheses, which increases with the period of wear and can reach 50% after 5 years. We find that the most commonly-cited reasons for rejecting a prosthesis are its weight, slow actuation speed, and poor durability, whereas the cosmetic appearance, ease of use and engineering value of the prosthesis contribute most towards its acceptance. We hypothesise that we can improve the acceptance rate of myoelectric prostheses by offering anthropomorphic devices which are lighter, more durable, and feature faster speeds of actuation. We research the anatomical structure of the human thumb, and construct four anatomically-accurate models using the DenavitHartenberg notation, which feature different axes orientations and placement. We obtain values for the angular range of motion of the thumb joints and quantify a range of motion of the human thumb in three-dimensional space by generating a point cloud of end-effector positions of the human thumb, and computing its volume by bounding it with alpha-shapes. We compare and contrast the four thumb models and extract a single optimisation target. We examine the function of the thumb when executing a number of ADLs, and reduce the range of motion required of a prosthetic/robotic thumb by 76% vis-a-vis the full range of motion of the human thumb. We produce an algorithm which generates and classifies reduced-DoF anthropomorphic models of the human thumb, with randomised axes orientation and spacing. Finally, we propose a twostep optimisation algorithm, utilising the Surface-Response Method, to converge to the optimum axes placement and orientation for family of thumb models.