Previous Abstract Return to Session C6 Next Abstract

Session C6: Terrestrial Signals of Opportunity-Based Navigation Systems

An Empirical Assessment of Indoor-Outdoor Localization Based on Signals of Opportunity from Multiple Systems
Albrecht Michler, Paul Schwarzbach, Jonas Ninnemann, Muhammad Ammad, Hagen Ußler, Oliver Michler, TUD Dresden University of Technology

Achieving seamless indoor-outdoor positioning remains a critical challenge in radio-based localization systems, particularly in transitional areas where satellite signal reception degrades due to limited sky visibility. Additionally, these areas often exhibit suboptimal geometry and limited coverage for indoor positioning systems. Moreover, during the transition from outdoor to indoor environments, radio systems require time to establish connections, introducing additional constraints on continuous localization solutions. We also address scenarios where none of the systems can compute a position solution due to insufficient measurements. In such cases, inter-system fusion becomes essential for maintaining localization functionality. This paper presents a empirical analysis of system availability and localization performance by employing a hybrid fusion of GNSS and terrestrial systems using a signal of opportunity approach.
Measurements were conducted at an automated bus test facility in Torgau, Germany, in October 2024. The focus is on the transition between outdoor areas and an indoor passageway. Both static and dynamic tests were performed to emulate realistic conditions and capture varying signal availability and characteristics. The experimental setup integrated multi-constellation GNSS with various commercial terrestrial positioning technologies, including Ultra-Wideband (UWB), Wi-Fi Fine Timing Measurements, Bluetooth Low Energy (BLE), and 5G NR localization systems. The characteristics and measurement procedures for each radio technology are introduced to highlight their strengths and weaknesses in positioning. For each system, we recorded available localization inputs (e.g., pseudoranges, two-way ranging) and referenced them against both an electronic tachymeter (Leica Viva TS15 Total Station) and a camera-based tracking system (Vicon Vero cameras in combination with Vicon Tracker software). Systems were time-synchronized using GNSS-disciplined NTP/PTP master servers. The dataset consists of 12 static measurement points and a dynamic trajectory, that traverses the measurement area, moving from outdoor to the indoor passageway and back to outdoor again. All measurement data will be published alongside the paper.
Methodologically, conventional and previously published state estimation methods are compared to assess the localization accuracy in the given scenario. This includes parametric least-squares and extended Kalman filter-based approaches, as well as a non-parametric filtering approach utilizing grid positioning within a Bayesian framework to fuse the heterogeneous positioning data. Furthermore, the approach facilitates the integration of additional process knowledge. This includes the known realization space, which is the area where the receiver can possibly be located under normal circumstances (e.g., a train is expected to be on tracks). Additional information can be extracted from known vehicle/pedestrian dynamics and motion models, environmental radio propagation models, or machine learning-based models that estimate common expected behaviors or other properties. Further information can easily be included, as long as this information can be mapped into a probabilistic representation.
The result of this study is a detailed analysis of data availability and system-specific accuracy metrics. This includes analyzing the geometric constellation of opportunistic signals, particularly in areas where the limited number of visible satellites and terrestrial observations prevent system-specific solutions from being computed. In addition, the accuracy of the surveyed observations and provided quantities for each localization system is evaluated by applying common evaluation metrics, such as ranging error and variance. Results demonstrate the feasibility of hybrid signal fusion for seamless indoor-outdoor positioning, with a particular focus on transitional areas between the two domains.
Positioning performance is evaluated using the root mean square error metric, based on the 2D and 3D geometric distances between the estimated positions and the reference positions. The evaluation is divided into three zones to account for different reception conditions: outdoor (full GNSS coverage), outdoor-indoor transition (limited GNSS plus local sensor signals of opportunity), and indoor (full local sensor coverage). While the transition zone is expected to have the worst positioning performance, the experimental validation is designed to examine this hypothesis.
In conclusion, we analyze the complementary benefits of combining GNSS and terrestrial positioning systems, considering the properties and measurement principles of each system, using an empirical dataset of heterogeneous radio systems. The given setup with four complementary terrestrial radio technologies allows for the fusion of GNSS, UWB, BLE, 5G NR, and Wi-Fi. The strengths of individual systems in terms of range, sensor density, accuracy, latency, and update rate are leveraged through heterogeneous positioning data fusion to enable an opportunistic positioning system. The main aim of the study lies in scenarios where no single system is able to estimate a position, and only an opportunistically fused solution provides a localization solution.

Previous Abstract Return to Session C6 Next Abstract