A multi-modal dataset for hybrid indoor positioning using Wi-Fi RSS, embedded inertial sensors, and CCTV images.

Indoor localization solutions are essential for delivering accurate location-based services in environments where Global Navigation Satellite Systems (GNSS) are ineffective. The Received Signal Strength Indicator (RSSI)-based, inertial sensors' reading, and captured images by CCTV camera leverages the existing indoor infrastructure to provide a cost-effective solution for indoor location determination. However, device heterogeneity, characterized by variations in hardware, sensors, and software between devices, poses a significant challenge, often degrading positioning accuracy and robustness. To this end, this paper presents a structured multimodal dataset that is aimed at fostering research in hybrid indoor localization. The data set was acquired in an indoor university corridor, making use of already existing infrastructure, i.e., Wi-Fi access points (AP)s, inertial sensors of smartphones, and a set of stationary CCTV cameras. The goal is to create a test facility for testing sensor fusion methods using Wi-Fi RSSI, embedded inertial sensors readings (accelerometer, gyroscope, magnetometer) along with visual information. Data were sampled over a 1 m resolution grid across the outer and second-inner edge of a 39 × 49-m corridor, with 10 repeated measurements per point in order to obtain a reliable measurement. Persons and door labels were identified using YOLOv8 based object detection and were annotated on CCTV images taken every 2 meters. Magnetometer readings are also available in the dataset to facilitate orientation and heading awareness. For localization accuracy, in the preliminary trials, weighted 3-Nearest Neighbours (W3KNN) was used in RSSI localization and the outputs were fused with image-based distance estimation result by employing KALMAN and Particle Filters. The dataset facilitates investigation in more advanced signal fluctuation modelling, vision-assisted localization, and machine learning-based trajectory estimation, as well as for the development and evaluation of robust hybrid indoor positioning systems. However, the dataset also has practical limitations, since the dataset is constructed in single building floor as well as limited number of utilized WiFi access point, CCTV cameras and a single held smartphone.