<title>Mobile robot navigation with vision-based neural networks</title>

Abstract

Mobile robot technology is spreading its use in the development of advance manufacturing systems. Methods of multi-sensory fusion data with vision, sonar and limit switches have been developed as the most flexible, but expensive approaches. Other approaches are more common such as buried wire AGV's. They decrease the cost of the mobile robot, but degrade the flexibility of the navigation system as well. This paper uses neural networks (NNs) with only one camera to obtain similar flexibility as the high cost approaches, but in a cost-efficient way. The NNs use translation and perspective information of features in images to determine the proper alignment and position of the mobile robot.