Neurofuzzy Inverse Jacobian Mapping for Multi-Finger Robot Hand Control

Fuzzy systems and models are useful for describing processes where the underlying physical mechanisms are not completely known and where a system behavior is understood in qualitative terms. Neurofuzzy systems have been employed in large number of intelligent based control systems and robotics, that is due to the ability to deal with large number of inputs and with the ability to learn and remember specific learned patterns. This paper investigates the employment of a neurofuzzy system for a multi-finger robot hand control and manipulation tasks. The approach followed here is to let a defined neurofuzzy system to learn the nonlinear functional relation that maps the entire hand joint positions and displacements to object displacement. This is done by avoiding the use of the Inverse Hand Jacobian, while observing the interaction between hand fingers and the object being grasped and manipulated. The developed neurofuzzy system approach has been trained for several object training patterns and hand postures within a cartesian based palm dimension. The paper demonstrates the proposed algorithm for a four fingered robot hand motion, where inverse hand Jacobian plays an important role in the hand dynamics and control.