PROJECT TITLE :

Enhanced Wi-Fi RTT Ranging: A Sensor-Aided Learning Approach

ABSTRACT:

Through the utilization of round-trip time (RTT) measurements, the fine timing measurement (FTM) protocol was developed with the intention of determining accurate ranging between Wi-Fi devices. The multipath propagation of radio waves, on the other hand, results in inaccurate timing information, which in turn reduces the performance of the ranging system. Throughout this investigation, we make use of a neural network (NN) to iteratively learn the one-of-a-kind measurement patterns that are seen in a variety of indoor settings. This allows us to generate improved ranging outputs from raw FTM measurements. In addition, the NN is educated based on an unsupervised learning framework, with the naturally accumulated sensor data obtained from users accessing location services serving as its primary source of information for its education. As a result, the amount of work required to collect training data is significantly reduced. The findings of the experiment demonstrated that it is possible to learn the pattern in raw FTM measurements and produce more accurate ranging results by collecting unlabeled data for only a short period of time. The proposed method resulted in a 47–50% reduction in the amount of error in raw distance measurements and a 17–29% reduction in the amount of error in well-calibrated ranging results, both of which required the collection of ground truth data. As a direct result of this, positioning errors were reduced by 17–30% in comparison to the scenario involving well-calibrated ranging.