The CityLab Testbed- large-scale Multi-technology Wireless Experimentation in a City Environment: Neural Network-based Interference Prediction in a Smart City Jakob Struye, Bart Braem, Steven Latr´e, Johann Marquez-Barja IDLab - University of Antwerp - imec Middelheimlaan 1, 2020 Antwerp, Belgium, firstname.lastname@uantwerpen.be
Testbed IoT is changing the way technology interacts with daily life. Testbeds provide an excellent means to validate Research & Development (R&D) results in realistic conditions CityLab
Requirement Experimentation Realism Reliability Multi-technology Mobility User support 通常，过去很多的实验床都是在被控制的环境下测试的，通过一些仿真模拟达到很不错的效果。但是，比如环境噪音就很难模拟，本文认为应该最大限度的真实环境 ，所以部署在一个小城市里。 可靠性，测试床必须提供高可靠性，使得使用者可以低延迟有效的远程控制节点，同时有自我修复性，使用者在网上就可以轻松操作了。 作者认为小城市应该面向所有技术，所以本文他们 802.11，sub-G Hz，同时对新技术有包容性。
CityLab-Architecture jFed is a Java-based framework for testbed federation
jFED
Outdoor and indoor unit
Locations and deployment
Use Case Predict the background noise 2.4GHz Wi-Fi
Experiment setup CityLab Nodes with COMPEX WLE900VX-7A network adapter, which contains a Qualcomm Atheros QCA9880 wireless chipset. spectral scan 本质上来说是一个简单的频谱分析仪
Data processing Use the FFT_eval tools to visualize the signal strengths across multiple channels at one point in time. Signal strength unrealistically high Discard outliers Savitzky-Golay filter
Neural Network Design Recurrent Neural Networks (RNNs) GRU layer,update gates and the reset gates.
result
result
Conclusion a novel smart cities experimental facility, which enables multi- technology experimentation in a realistic smart cities context, at a large scale After filtering our measurements, we revealed a clear daily pattern in the interference levels in one case. Designed and tuned an efficient neural network to predict the interference.
Related work SmartSantander in Santander, Spain Ubiquitous Oulu Smart City testbed in Oulu, Finland
Interaction for experimentation
Web interface
Oulu Smart City
Run with us system
Software stack