Propagation Path loss model based on Environmental Variables

• Published in: 2020 12th International Conference on Information Technology and Electrical Engineering (ICITEE) pp. 368 - 373
• Main Author: Bolli, Sridhar
• Format: Conference Proceeding
• Language: English
• Published: IEEE 06-10-2020
• Subjects: ANFIS; ANN; Cognitive Radio; Computational modeling; Data models; Fuzzy logic; Image segmentation; Indexes; Loss measurement; path loss; Roads; Urban areas
• DOI: 10.1109/ICITEE49829.2020.9271731

We have developed a path-loss model that includes environmental variables. We take a sizeable 2-dimensional satellite image of 4 cities, namely Hyderabad, Mumbai, Chennai, New Delhi, and then divide the large 2d image into many smaller images. Then we perform image segmentation using the Maximum likelihood algorithm on each smaller image. Segmentation separates the image into separate areas comprising of pixels with identical qualities. After that, we develop three different 11 input path loss models based on Fuzzy logic, Artificial Neural Network (ANN) and Adaptive Neuro-Fuzzy Inference System (ANFIS), respectively. Input Parameters to all these three path loss models were %building, %road, %plain, %water, %trees, transmitter terrain height, receiver terrain height, the distance between receiver and transmitter, average clutter height, transmitter frequency, and transmitter height. The output of all the above three models is a path loss. We acquired receiver power levels data in a driving test through different routes in all four cities. We compared measured path-loss values for each route with the predicted values obtained with ANN(with image segmentation), ANFIS(with image segmentation), FCM(with image segmentation), ANFIS(without image segmentation), and empirical path loss models. We measured each path-loss model's accuracy with RMSE (root mean square error) obtained between the predicted & measured path loss values. This paper found that ANFIS(with image segmentation) path-loss model has an RMSE of 2.16 dB, the lowest RMSE among all the considered path-loss models.