Road and Intersection Detection Using Convolutional Neural Network

Abstract

It is an important task to detect the direction in which the autonomous robot can move. The robot can autonomously move to its destination following the information on how many times it detects the intersections and turns. In this paper, we use a convolutional neural network (CNN) to simultaneously recognize both the direction along the road and the intersection. The CNN detects the directions through the map image built using the scan data from a two-dimensional laser range finder (2D LRF). We show that the robot is able to make an autonomous movement along the road until it detects the intersection where it should turn.