Visible light‐based indoor localization using k‐means clustering and linear regression

Visible light positioning techniques employing received signal strength (RSS)–based fingerprints are becoming popular and ubiquitous. However, RSS is more susceptible to signal degradation and environmental changes resulting in location inaccuracies. To minimize these limitations, clustering in conjunction with linear regression is applied to RSS database made up of light intensity variations of light emitting diodes. Optimum cluster size is determined and trained clusters are exploited for location assessment by curtailing the difference between the database readings and cluster centroids. Regression is then applied on the clustered data, which partitions it further and helps refining the results. Simulation results of the proposed algorithm dictate a significant improvement in location estimation accuracy of up to 40 cm in an indoor environment with the dimensions of 5 m × 5 m × 4 m and exhibit superior performance than many state‐of‐the‐art RSS‐based methods. A novel visible light based indoor localization system has been developed utilizing received signal strength (RSS) and machine learning approaches. Results show that signal degradations and other environmental changes resulting location inaccuracies in RSS can be minimized using machine learning techniques.