Towards Modelling Autonomous Mobile Robot Localization by Using Sensor Fusion Algorithms

Abstract

Autonomous Mobile Robot (AMR) is widely used in a variety of applications. This paper describes an early experiment towards modelling a low-cost and robust centimetre-level localization for mobile robots in crowded indoor and outdoor environments. While a wide range of methods have been developed and tested on high-end hardware in autonomous vehicles, the work utilizes multiple sensor information to achieve robustness with different types of mobile robots. The application can be used by any group or organization, especially the frontliners, in managing the COVID-19 pandemic. Different Simultaneous Localization and Mapping (SLAM) algorithms, such as GMapping, Google Cartographer and Hector SLAM, are used to achieve better localization. Sensor fusion strategy is applied for these SLAM packages using Real-Time Kinematic (RTK) positioning, a precise Global Navigation Satellite System (GNSS)-based sensor, by applying both Extended Kalman Filter (EKF) and Unscented Kalman Filter (UKF) to estimate position, velocity and attitude (PVA). The performance of the proposed algorithm will be compared against the benchmark algorithm using different sets of data in crowded places in various settings.