Seamless indoor/outdoor environment aware localization through indoor 5G NR & GNSS

Global navigation satellite system (GNSS) positioning is arguably the standard for localizing mobile robots in outdoor spaces. Localization in indoor spaces employs many radio-based systems, such as 5G, Wi-Fi, and Bluetooth. However, robots that traverse between outdoor and indoor areas and vice versa often rely on simultaneous localization and mapping techniques, which require accurate pre-built maps of the environment for global localization. Such maps may not always be available, making external positioning systems preferable for seamless localization across heterogeneous indoor and outdoor environments. This work proposes a novel seamless indoor-outdoor localization approach based on external radiobased sensor systems, including GNSS for outdoor areas and 5G for indoor localization. An environment-aware Markov decision process and an extended Kalman filter govern the fusion of the data to ensure the appropriate integration of external radiobased sensors, providing a seamless localization estimate across heterogeneous areas. The proposed framework is thoroughly evaluated in two different environments through real-world experiments and achieves a median localization error of 1.2 m while the mobile robot traverses indoor and outdoor areas, delivering significant performance improvements over baseline solutions and raw sensor readings.