RadioResUNet: Wireless Measurement by Deep Learning for Indoor Environments

This paper introduces a wireless measurement method by deep learning that trains raytracing data and inferences wireless conditions in indoor environments. For indoor wireless measurement by deep learning, we prepare a large-scaled training dataset consisting of indoor interior models imitating various real indoor environments and radiomap images generated by raytracing radio wave propagation in the indoor interior models. In addition, we propose a deep learning model for inferencing the indoor wireless conditions, namely RadioResUNet, that is based on a convolutional neural network (CNN). To evaluate the deep learning model, we compare the proposed deep learning model of RadioResUnet with the existing deep learning model of RadioUNet. To verify the feasibility and effectiveness of the proposed RadioResUNet, in addition, we compare the received signal strength results by deep learning measurement with those by actual measurement in the indoor environment and show the effectiveness and challenges of the deep learning model of indoor wireless measurement.