Towards Context-Aware Indoor Positioning for IIoT Using Dnn-Based Fingerprinting with AP Selection

Abstract

Emerging Industrial Internet of Things (IIoT) ap-plications-such as asset tracking, worker safety, and autonomous indoor navigation-demand accurate and adaptive positioning systems capable of operating in dynamic indoor environments. Classical deep neural network (DNN)-based fingerprinting methods often assign equal weight to all access points (APs), without ranking or weighing them based on signal relevance, which can compromise localization accuracy and increase computational burden, especially in complex settings. To overcome these limitations, we develop a DNN-based fingerprinting framework that incorporates a dynamic, context-aware AP selection module to improve both positioning accuracy and efficiency. By ranking and weighing APs based on their relative contribution to localization performance, the proposed method enhances adaptability and precision. We validate the framework using a hybrid dataset composed of received signal strength indicator and time-of-flight measurements. Experimental results show that our proposed method yields significantly better positioning accuracy than classical approaches. This work establishes a foundation for context-aware indoor localization systems designed to meet the performance and reliability requirements of IIoT applications.