Geometry-Aware & Lightweight UWB Positioning for Vehicle Applications

Sammanfattning

Ultra-wideband (UWB) positioning is increasingly used in automotive applications, such as keyless entry and access control, where accurate and reliable localization of devices in a vehicle’s immediate vicinity is required. In this setting, anchors are mounted at fixed locations on the vehicle body, the region of interest is limited, and non-line-of- sight (NLOS) propagation caused by the vehicle itself is common. These characteristics distinguish near-vehicle positioning from general indoor localization and motivate the use of algorithms that explicitly exploit problem structure while remaining computationally lightweight, as such systems are typically deployed on resource-constrained embedded platforms. This thesis investigates range-based UWB positioning methods for near-vehicle scenar- ios, with a focus on geometry-aware design and predictable computational cost. A set of established positioning algorithms is implemented and evaluated, including iterative least squares, weighted least squares, and extended Kalman filter (EKF)– based approaches. Several lightweight NLOS mitigation strategies are considered, ranging from geometric consistency checks to simple weighting schemes derived from known vehicle geometry. Temporal filtering is applied both as post-processing and by directly incorporating range measurements into filtering frameworks. The algorithms are evaluated using a combination of simulation and real-world experi- ments conducted around a stationary vehicle equipped with four UWB anchors. Position- ing accuracy is assessed across multiple motion patterns, while computational complexity is analyzed using worst-case floating-point operation counts to evaluate feasibility on em- bedded hardware. By comparing algorithm variants under consistent assumptions, the study highlights trade-offs between accuracy, robustness to NLOS conditions, and com- putational cost. The results show that exploiting the structured geometry of the near-vehicle setting enables simple, range-based algorithms to achieve reliable positioning performance with- out resorting to complex signal-level processing or data-driven methods. The findings suggest that geometry-aware, lightweight approaches provide a practical basis for real- time UWB positioning in automotive applications.