Joint control of ROBOKER arm using a neural chip embedded on FPGA

Abstract

This paper presents implementation of a neural chip to proceed neural processing of the radial basis function (RBF) network. RBF network along with a primary PD controller is trained in on-line fashion. Radial basis function network processing is embedded on a field programmable gate array(FPGA) chip to achieve real-time control. To enable nonlinear function calculation, a floating point processor is designed to allow assembly programming for learning algorithm. Other necessary hardware modules for control purposes are also designed and implemented. A humanoid robot called the ROBOKER with two arms of 6 degrees-of-freedom each is controlled. Joint angles of the ROBOKER arms are controlled and tracking performances by the neural chip are compared with those by PD controllers.