A neural network approach to indoor mobile robot localization

Abstract

In order to improve the real-time performance and accuracy of localization for mobile robot in indoor environment, a neural network data fusion approach is proposed to eliminate the affection caused by errors from environment or measurements. In the approach, the odometry data are firstly obtained by calculating the collected encoder data through the Dead Reckoning (DR), then we fuse the odometry data and the lidar data by inputting them into a three-layer neural network. Experimental results show that the trained network improved the robot localization performance and its position accurate is within 6cm with good real time response.