Enhancing Multi-Floor Indoor Localization Accuracy Using Fingerprint-Based Dynamic $k-\text{NN}$ Approach

Accurately localizing indoor spaces with multiple floors presents a significant challenge, yet it is crucial for a range of applications from smart buildings to emergency response systems. This study presents a novel approach to multi-floor indoor positioning using Wi-Fi received signal strength (RSS) fingerprinting. Our method integrates a novel voting scheme for floor differentiation and a dynamic “valid neighbor” selection technique for user localization, both implemented within a <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\boldsymbol{k}$</tex> -nearest neighbors (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\boldsymbol{k}$</tex> -NN) framework. The approach was tested and validated using both sparse and dense fingerprint datasets collected across three floors of an engineering building, featuring various room sizes and layouts. Enhanced by additional components, our voting scheme achieved complete accuracy in floor differentiation for the dense dataset, improving from 99.82% to 100%, and increased accuracy in the sparse dataset from 94.93% to 97.46%. Additionally, using the dynamic selection approach, we reduced the mean localization error from 3.04 meters to 2.87 meters in the sparse dataset, and from 1.65 meters to 1.53 meters in the dense dataset.