Indoor RSSI Prediction using Machine Learning for Wireless Networks

We consider the study of received signal strength indication (RSSI) prediction in an indoor room environment using a small set of actual measurement data. The RSSI prediction in a test environment is important in a network planning strategy. Traditional models are based on either empirical or deterministic models, which are time-consuming depending on many factors such as the room structure, obstacles, and many more. In this paper, we investigate any simple machine learning model like an artificial neural network (ANN) or linear regression model that can do this RSSI prediction and the estimation of environmental-related parameters that affect the prediction of RSSI. We assume that some parameters like transmit power, antenna height, wall material properties are kept fixed. We illustrate the RSSI prediction performance in terms of mean squared error (MSE) and mean absolute error (MAE) for a dataset with 1030 data points collected from the test environment. The path loss exponent of our test environment is estimated as 1.97.