RSSI-Based Indoor Distance Estimation in Wi-Fi IoT Application Using AI Approaches

In many Internet of Things (IoT) applications, device tracking is required to improve localization-based services. Distance estimation is a key component of localization methods. Accurate distance estimation helps Wi-Fi access point's providers, to provide better quality of services (QoS) to connected devices. Distance estimation determines a device's precise location in an environment. In this work, three different scenarios are considered. The work proposes distance estimation using Euclidean distance and the received signal strength indicator (RSSI). This approach eliminates the need for three access points required in the triangulation method. Euclidean distance is measured between access points and devices spread throughout the indoor environment. Random forest (RF) algorithm, K-nearest neighbor (KNN), Gaussian process (GP), and ant colony optimization (ACO) are used to train the models to correlate Euclidian distances and the RSSI of connected devices. The performance is measured using mean absolute error (MAE). The optimum performance improvement obtained in Scenario 1, for RF, is 41.87%; in Scenario 2, for KNN, it is 37.98%; and in Scenario 3, for RF, it is 56.97%, with respect to earlier reported work. ACO achieves a 78.61% improvement in Scenario 1, 88.57% in Scenario 2, and 50.90% in Scenario 3 over the reported work.