RLRL: Robust Low-cost robot localization on diverse road surfaces via wheel encoder
Jeonghyeon Yoon, Seungku Kim · Measurement · 2025
Recent research on autonomous mobile robots has garnered significant attention. Used in diverse sectors such as smart factories, agriculture, and commerce, autonomous mobile robots have the potential to replace human roles. For efficient operation and goal achievement, it is crucial for these robots to accurately determine their position. Most studies on robot localization use expensive sensors like LiDAR and cameras to ensure localization accuracy within 1 m. This paper aims to perform robot localization at a lower cost using only inexpensive wheel encoders. The proposed robot localization technology applies a model that classifies road surfaces and performs localization based on the classified road surface type. Experimental results indicate that the proposed robot localization technology achieved an average performance improvement of 42 % compared to conventional wheel odometry methods in real-world experiments of varying distances. In the 1700 m experiment, the wheel odometry scheme exhibited an error of 87.60 m, whereas the proposed robot localization technology demonstrated a reduced error of 41.66 m. Furthermore, it operates within approximately 7.2 million floating-point operations, enabling real-time localization even on devices with low computational power.