Opportunistic WiFi Spectrum Reuse for Car Density Estimation

Signal reuse for multiple purposes is a way to increase spectrum utilization. In this paper, we leverage the WiFi signals of opportunity for the sensing purpose. The spectrograms derived from WiFi downlink (DL) signals reflected from cars are used as fingerprints to efficiently infer car density in parking lots. To achieve this, experimental measurements were conducted in a real outdoor environment to probe the reflected WiFi signals from targets (cars), and the collected datasets are employed for density estimation. The estimator combines hybrid convolutional neural network (CNN) and support vector machine (SVM) for classification, along with least-square estimate (LSE) for interpolation. The probed signals are influenced by many factors, such as the number of WiFi users and data traffic, thereby degrading the estimation accuracy. To address these challenges, we propose an uplink-downlink (UL-DL) WiFi identification and separation technique using the least absolute shrinkage and selection operator (LASSO) technique, without requiring coordination with WiFi access points. Compared to the estimation using a mixture of UL-DL WiFi signals, the simulation results demonstrate that the proposed method achieves significant improvement in estimation accuracy.