An End-to-End Deep Learning Model based on Channel Impulse Response Measurement for 2D Indoor Positioning

for  a long time, numerous methods have been available to estimate the position in an indoor environment using a wireless signal. in this work, we are considering the channel impulse response (cir) signal that is retrieved from the base stations (bss). the cir signal has more position information compared to other signals. we developed an ai model called spatialn et to perform indoor positioning. the model retrieves input data from 18 bss and performs feature extraction individually for each bs to recover the unique and important features. then, an attention block is utilized to focus only on the important parts of the extracted features in the dataset. the spatialnet model is used to predict the 2d position of the user equipment (ue). thus, the output of the spatialn et model is the position of ue in x and y coordinates. a cnn model is developed to be compared with the spatialnet. from the training and testing results, the spatialnet achieves the best performance in terms of mse, mae, and smape metrics compared to the cnn model with value of 0.1907, 0.3207, 0.7524, respectively.