Neuro-Kinematics Based Dexterous Robotics Hand Force Optimization

The complexity of computing appropriate distribution of manipulation forces among fingers of a four-fingered robot hand defined in a dynamically Task-Space coordinate task is addressed. Finger-object interactions are modelled as point frictional contacts, consequently, the system is indeterminate. Hence, an optimal solution does necessitate for controlling forces acting on a grasped object. A fast and efficient method for computing optimal grasping and manipulation forces is presented based on a quadratic optimisation formulation, where computation has been based on using the nonlinear factual model of contacts. Furthermore, in order to achieve grasping while in motion, the hand inverse Jacobian has to be intensively computed, consequently, we investigate an efficient approach of employing an artificial neural network for the multi-finger robot hand in which the object motion is defined in. The approach followed here is to let an ANN to learn the nonlinear inverse kinematics functional relating the hand joints positions and displacements to object displacement. This is done by considering the inverse hand Jacobian, in addition to the interaction between hand fingers and the object being grasped and manipulated.