Mobile robot localization using the Hough transform and neural networks

For an autonomous mobile robot to navigate in an unknown environment it is essential to know the location of the robot on a real-time basis. Finding position and orientation of a mobile robot in a world coordinate system is a problem in localization. Dead-reckoning is commonly used for localization, but position and orientation errors from dead-reckoning tend to accumulate over time. The objective of the paper is to develop a feature-based localization method that uses the Hough transform to detect wall-like features in the environment based on sonar range data. Although the Hough transform is an effective method for detecting lines and curves from noisy data it has the drawback of being sensitive to discretization resolution. Results of line detection using various discretization resolutions and using a winner-take-all neural network are presented and discussed. The results indicate that the Hough transform method is able to reliably recognize wall-like features from noisy sonar data, but the accuracy of detected features is dependent heavily on the choice of resolution of parameter discretization