Learning based Wi-Fi RTT Range Estimation

In Wi-Fi RTT range estimation, accuracy is the most critical issue. But current estimated values using Wi-Fi RTT with 802.11mc FTM protocol are often randomly far away from the true range. These inaccuracies and fluctuations make it difficult to estimate the distance of mobile devices and Wi-Fi access points needed for indoor location-based services. In this paper, we present learning-based system model to get generalized probabilistic distribution. We made a deep learning model using existing measured range values on each certain range as training data. To improve accuracy, we used multiple correlated parameters detected with 802.11mc FTM. We verified the performance of our model using real test data. It is shown that it can guarantee the stability with high accuracy for true range estimation. Our system can be used as a base framework for other various situations or more learning algorithms to enhance development efficiency.