Single Base Station mmWave Radio Positioning, Mapping, and SLAM

Sammanfattning: Fifth-generation (5G) communication systems in Frequency Range 2, operating above 24 GHz and utilizing mmWave signals, showcase distinct properties that open up new possibilities in positioning, mapping, and simultaneous localization and mapping (SLAM). The combination of large bandwidth, extensive antenna arrays, and high carrier frequency results in geometric-based signals, and unprecedented delay and angle resolution. These enable the system to resolve multipath components, providing high-accuracy geometric information among the user equipment (UE), the base station (BS), and the environment. These high-accuracy geometric information allows for highly accurate UE positioning, environment mapping, and SLAM, all achievable using a single BS. While numerous studies have delved into the single BS positioning and mapping problem using snapshot measurements, a significant portion of them remains confined to theoretical analyses with many simplified assumptions. Real-world experimental validation is scarce, particularly in scenarios involving a commercial 5G BS. Additionally, while diffuse multipath contains valuable geometric information, it is often treated as a perturbation or fails to acknowledge that diffuse multipath signals may originate from the same source landmark, leading to information loss. When extending positioning and mapping to a SLAM problem by tracking the UE over time, a radio SLAM problem emerges, posing the primary challenge of effectively addressing the data association (DA) problem. It is these research gaps and challenges that drive the motivation behind this thesis. Within this thesis, [Paper A] and [Paper B] address the radio SLAM problem, with [Paper A] additionally exploring the utilization of diffuse multipath. In [Paper A], we adopt an end-to-end approach to address the radio SLAM problem, presenting a comprehensive framework for SLAM. This includes the introduction of a random finite set (RFS)-based SLAM filter designed to overcome the DA challenge inherent in radio SLAM, along with a method to effectively leverage all paths originating from the same landmark. Meanwhile, an efficient alternative RFS-based SLAM filter, designed for real-time implementation, is proposed in [Paper B] as a counterpart to the solution presented in [Paper A]. In [Paper C], we focus on experimental validation of positioning and mapping with a single BS, showcasing the practical feasibility while uncovering existing gaps between theoretical expectations and real-world implementation. [Paper D] delves into the fusion problem involving mapping and SLAM results from various sources, presenting an RFS-based fusion solution.