Implementation of RAM Based Neural Networks on Maze Mapping Algorithms for Wall Follower Robot

Abstract

A wall follower robot will recognize its environment by knowing the position of the surrounding wall. The goal is that the robot has a good safe navigation system without damaging the walls. For navigation systems, robots use the Maze Mapping method with the left-hand rule, while to study the wall distance pattern is a RAM-based artificial neural network method. The neural network has 3 RAM nodes to process the received environmental patterns. The left sensor, the right sensor, and the front sensor have 8 bits of the input pattern, the pattern will be optimized into 4 bits stored in the RAM node, so that the computing process in the robot becomes faster and simpler [1][4]. RAM discriminator design has 3 RAM nodes. Each RAM node has 4 bit words (x = 4), with a total of 12 input vector bits (n = 12). Each RAM discriminator will receive 48 binary input patterns. The results of environmental polo testing, the wall follower robot can complete the labyrinth path which has 8 intersections and a dead end with a distance of 595 cm, then the next 4 intersections with a distance of 360 cm.