Radio Wave Propagation Prediction Based on Path Sequence Regression

In this paper, a new method for radio wave propagation prediction based on path sequence regression is proposed. Firstly, in single antenna and single cell scenarios, the features of antenna, receiver and surrounding environment are extracted, and the data set is obtained. Then we randomly select the neighboring receiving points for signal diffusion, and an undirected connected graph is constructed to simulate the random propagation of signals. In the undirected connected graph, the shortest path tree from the antenna to the receiving points can be obtained. Next, we extract the features in each receiving point in the shortest path to construct the sequence of feature vectors. Finally, we use the sequences of feature vectors to train the GRU-MLP network to predict the reference signal receiving power (RSRP) of the receiving points.