A Machine Learning Approach for Wi-Fi RTT Ranging

Abstract

The following paper, presents a supervised, deep-learning, neural network approach for range estimation based on IEEE 802.11 wireless local area network (WLAN) round-trip timing (RTT) measurements. The range estimation accuracy is compared against a standard, time-of-arrival (TOA), maximum-likelihood estimation (MLE)-based range estimation. The deep-learning approach is based on a “Siamese”, artificial neural network (ANN), which was trained using both indoor channel simulation, as well as actual channel measurements collected in a real, indoor office environment. Both the MLE and ANN range estimators were tested using real-channel measurements and the estimation accuracy was analyzed using “ground-truth” information collected using a LiDAR system. It is shown that the ANN-based approach outperforms the accuracy achieved by the classical MLE approach.