Magnetic field-based mobile robot localization in repetitive environments

Abstract

With the rapid development of mobile robots, a robust localization algorithm that could enable the robots to obtain accurate location within various kinds of environments is expected. However, the widely-used Global Navigation Satellite System (GNSS) may be unstable or even denied under certain circumstances. Consequently, new solutions need to be developed. Over the past several decades, place recognition and localization has proven to be a promising solution. Current robot localization methods can be broadly divided into two categories: infrastructure-based and infrastructure-free. For infrastructure-based methods, infrastructures such as WiFi, UWB, and QR codes need to be pre-installed in the environments to provide unique location information. However, the installation and maintenance of the infrastructures cost both resources and time. In contrast, infrastructure-free methods are independent of these infrastructures and only rely on the environment information collected by the onboard sensors to localize the robot. However, there are still challenging scenarios for infrastructure-free methods. Repetitive and symmetric environments, such as the corridors, carparks, and warehouses, usually have highly similar structures and few unique features. Infrastructure-free localization methods can be inaccurate or even fail in such environments. Among the infrastructure-free localization methods, magnetic field-based localization has proven to be a viable alternative to the LiDAR-based or vision-based localization due to its pervasiveness and ease of implementation. Therefore, this report will explore to solve the infrastructure-free mobile robot localization problem in repetitive and symmetric environments based on magnetic field information and the proposed improved particle filter algorithm.