The CityLab testbed — Large-scale multi-technology wireless experimentation in a city environment: Neural network-based interference prediction in a smart city

摘要

Smart cities form an important new paradigm for future cities, where technology assists people, local economy and government. For smart cities to mature, it is crucial to enable experimental evaluation of current and new technologies, in realistic conditions. This paper contributes to the network testbeds domain by introducing the CityLab testbed, where researchers can experiment with a variety of smart city network technologies in parallel, including IEEE 802.11, IEEE 802.15.4, and sub-GHz protocols -on bare metal hardware enabling full software flexibility. As a second contribution, one aspect of realism, interference, is shown to be measured and predicted, based on data from the CityLab deployment. Specifically, we predict interference one hour into the future using a neural network based on a Gated Recurrent Unit. Compared to a naive predictor, the neural network is over 6.5 times as accurate.