Lightweight Robot Moving Distance Estimation via a Road Surface Classification

In this paper, we introduce a robot moving distance estimation system using wheel encoder sensors. While wheel encoder sensors are cost-effective and relatively straightforward for motion estimation, they are highly sensitive to errors affected by external factors. To address this problem, we employ deep learning technology to classify road surface and estimate moving distance with robots. The proposed system demonstrates approximately 60 percent improvement in terms of moving distance estimation error compared to the traditional wheel odometry technique.