Indoor Positioning Systems (IPS) traditionally rely on odometry and building infrastructures like WiFi, often supplemented by building floor plans for increased accuracy. However, the limitation of floor plans in terms of availability and timeliness of updates challenges their wide applicability. In contrast, the proliferation of smartphones and WiFi-enabled robots has made crowdsourced radio maps – databases pairing locations with their corresponding Received Signal Strengths (RSS) – increasingly accessible. These radio maps not only provide WiFi fingerprint-location pairs but encode movement regularities akin to the constraints imposed by floor plans. This work investigates the possibility of leveraging these radio maps as a substitute for floor plans in multimodal IPS. We introduce a new framework to address the challenges of radio map inaccuracies and sparse coverage. Our proposed system integrates an uncertainty-aware neural network model for WiFi localization and a bespoken Bayesian fusion technique for optimal fusion. Extensive evaluations on multiple real-world sites indicate a significant performance enhancement, with results showing ∼ 25% improvement over the best baseline.

Wi-Fi radiomaps offer a convenient solution to position users in GNSS denied areas, however the process of obtaining them through manual surveys is expensive and time consuming, while the use of crowdsourced uncalibrated data is noisier and lacks coverage. In this work we propose to calibrate the crowdsourced data using an optimization framework based on GraphSLAM using GPU processing, in order to automatically cluster the data into floors and perform 2D graph optimization and alignment, combined with opportunistic map matching when floor plans are available. Using our approach we are able to generate radio maps of multi floor buildings, processing data from thousands of user and millions of fingerprints. In our experiments we are able to produce a radiomap with a floor accuracy above 98% and median positioning error of 7.63 m in a 14 floor shopping mall (Joy City) in Beijing, China using single shot WKNN for localization.

Global deployment of Wi-Fi radiomaps is the key to high-accuracy, low-cost positioning systems that can provide indoor positioning at scale. Data-driven approaches to automate the creation of these radiomaps using unlabelled data are becoming increasingly popular. However, many systems rely on highly accurate indoor maps and low-noise crowdsourced trajectories. For deployment at scale, the concerns of inaccuracies in both the map and the data domains must be considered. In this research, we propose a 2-stage approach for calibration-free radiomap construction consisting of unsupervised trajectory alignment followed by a map matching optimisation stage. We evaluate the crowdsourced radiomap quality by utilizing an extensive ground truth data set consisting of thousands of estimates spanning 26 floors in 4 venues. We run positioning based on a singleshot Wi-Fi positioning algorithm (WKNN) and a particle filter-based recursive state estimation algorithm (PF). These algorithms achieve a median positioning error of 2.2 m and 1.3 m in an office environment, respectively. In larger mall environments, the average median errors are 8.1m (WKNN) and 4.6 m (PF).

Traditional Indoor Positioning Systems (IPS) use odometry, WiFi, and often building floor plans for accuracy. However, floor plan limitations have shifted attention to crowd-sourced radio maps, popularized by smartphones and WiFi-integrated robots. These maps pair locations with Received Signal Strengths (RSS) and reflect movement patterns similar to floor plans. Our research explores using radio maps as an alternative to floor plans in IPS. We've developed a new framework that combines an uncertainty-aware neural network for WiFi positioning with a Bayesian fusion method. Testing in real-world scenarios showed about a 25% performance increase compared to the leading baseline.

K-Neares Neighbors (KNN) and its variant weighted KNN (WKNN) have been explored for years in both academy and industry to provide stable and reliable performance in WiFi-based indoor positioning systems. Such algorithms estimate the location of a given point based on the locality information from the selected nearest WiFi neighbors according to some distance metrics calculated from the combination of WiFi received signal strength (RSS). However, such a process does not consider the relational information among the given point, WiFi neighbors, and the WiFi access points (WAPs). Therefore, this study proposes a novel Deep Neighborhood Learning (DNL). The proposed DNL approach converts the WiFi neighborhood to heterogeneous graphs, and utilizes deep graph learning to extract better representation of the WiFi neighborhood to improve the positioning accuracy. Experiments on 3 real industrial datasets collected from 3 mega shopping malls on 26 floors have shown that the proposed approach can reduce the mean absolute positioning error by 10% to 50% in most of the cases. Specially, the proposed approach sharply reduces the root mean squared positioning error and 95% percentile positioning error, being more robust to the outliers than conventional KNN and WKNN.

In recent years WiFi became the primary source of information to locate a person or device indoor. Collecting RSSI values as reference measurements with known positions, known as WiFi fingerprinting, is commonly used in various positioning methods and algorithms that appear in literature. However, measuring the spatial distance between given set of WiFi fingerprints is heavily affected by the selection of the signal distance function used to model signal space as geospatial distance. In this study, the authors proposed utilization of machine learning to improve the estimation of geospatial distance between fingerprints. This research examined data collected from 13 different open datasets to provide a broad representation aiming for general model that can be used in any indoor environment. The proposed novel approach extracted data features by examining a set of commonly used signal distance metrics via feature selection process that includes feature analysis and genetic algorithm. To demonstrate that the output of this research is venue independent, all models were tested on datasets previously excluded during the training and validation phase. Finally, various machine learning algorithms were compared using wide variety of evaluation metrics including ability to scale out the test bed to real world unsolicited datasets.

Floor determination is one of the challenges in indoor positioning research. To date indoor positioning solutions for floor determination have been mainly based on either Fingerprinting [2] or RF-ID [3]. While these solutions have been able to locate persons or equipments accurately even in multi story buildings, these can not be considered as time and cost efficient solutions. Therefore, in this research we present new Wi-Fi based indoor positioning algorithms in order to accommodate the need for limited resources solution in terms of deployment time and cost. While finger-printing and RF-ID based solutions have been targeting sub-meter position accuracy, this research focuses on floor determination only. In addition, in this research we have highlighted two possible approaches by using available Wi-Fi signals for floor determination.