Microcontroller Based Neural Network Controlled Low Cost Autonomous Vehicle

Abstract

In this paper, design of a low cost autonomous vehicle based on neural network for navigation in unknown environments is presented. The vehicle is equipped with four ultrasonic sensors for hurdle distance measurement, a wheel encoder for measuring distance traveled, a compass for heading information, a GPS receiver for goal position information, a GSM modem for changing destination place on run time and a nonvolatile RAM for storing waypoint data; all interfaced to a low cost AT89C52 microcontroller. The microcontroller processes the information acquired from the sensors and generates robot motion commands accordingly through neural network. The neural network running inside the microcontroller is a multilayer feed-forward network with back-propagation training algorithm. The network is trained offline with tangent-sigmoid as activation function for neurons and is implemented in real time with piecewise linear approximation of tangent-sigmoid function. Results have shown that upto twenty neurons can be implemented in hidden layer with this technique. The vehicle is tested with varying destination places in outdoor environments containing stationary as well as moving obstacles and is found to reach the set targets successfully.