Hardware realization of neural network based controller for autonomous robot navigation

Abstract

Recently soft computing techniques and artificial intelligent techniques such as fuzzy logic Artificial Neural Network are widely used for variety of systems, namely, controller architecture, pattern learning, navigation etc. This paper describes an alternative hardware solution realized on FPGAs for autonomous mobile robot to avoid obstacles and plan path to reach the target. Pipelined based Artificial Neural Network based controller architecture is proposed using FPGA. The proposed ANN algorithm is able to perform the task for unstructured environment and diverse environments. Simulation and hardware implementation has been done by using Xilinx ISE simulator targeted on Virtex-IV kit. Experimental study shows that proposed approach obtains 357.5 MHz clock frequency which shows improved performance when compared with state-of-art techniques. Similarly, proposed approach shows a significant performance improvement in terms of power consumption.