A multi-sensor fusion approach for centimeter-level indoor 3D localization of wheeled robots

Abstract Accurate three-dimensional positioning is fundamental for the safe movement of wheeled robots across multiple floors. However, achieving precise positioning of mobile systems in multilevel scenarios using low-cost sensors remains a significant challenge. This paper proposes a low-cost three-dimensional positioning method based on event-triggered extended Kalman filtering (EKF), which integrates data from a barometer, inertial measurement unit, and encoder. Firstly, during multilevel movement, sensors collect acceleration and barometric pressure data, which, combined with the EKF-target compensation (EKF-TC) fusion algorithm, achieve centimeter-level positioning of the wheeled robot’s height. Secondly, based on MEMS data, the integration of the dead reckoning algorithm reduces the horizontal positioning error of the wheeled robot over small obstacles. Finally, four horizontal and three vertical motion scenarios were designed to test the three-dimensional positioning capabilities of the wheeled robot moving between floors. Experimental results show that the cumulative vertical error of the wheeled robot moving across multiple floors accounts for approximately 0.6% of the total height of 93.76 m, with an average height positioning error of only 1.7 cm per floor. In summary, the method proposed using low-cost sensors and simple, stable fusion algorithms, offers a technical solution for indoor three-dimensional positioning with centimeter resolution for wheeled robots.