Deep Learning Localization with 2D Range Scanner

Abstract

In recent years, the use of 2D laser range scanners is increasing in industrial products, thanks to decreasing cost of this kind of devices and increasing accuracy. Nevertheless, the localization estimation of the moving objects (vehicles, robots, drones and so on) between consecutive laser range scans is still a challenging problem. In this paper, we explore different neural network approaches, using only a 2D laser scanner to address this problem. The proposed neural network shows promising results in terms of average accuracy (about 1cm in translation and 1° in rotation of Mean Absolute Error (MAE)) and in terms of overall used parameters (less than one hundred thousand), being an interesting method that could complement or integrate traditional localization approaches. The proposed neural network processes about 8000 pairs of compacted scans per second on Nvidia Titan X (Pascal) GPU.