Digital Twins for Smart Cities: Case Study and Visualisation via Mixed Reality

Abstract

Digital twins is an increasingly valuable technology for realising smart cities worldwide. Visualising this technology using mixed reality creates unprecedented opportunities to easily access relevant data and information. In this paper, a digital twins-based system is designed to visualise information from a city’s street lighting system. Data is obtained in two ways: from measured parameters of a miniature model street light in realtime, and from real Durham street lighting. Machine learning is used to maximise the efficiency of purchasing electricity from the grid, and to forecast appropriate adaptive street light brightness levels based on city’s traffic flow and solar irradiance. An application designed in Unity Pro is deployed on a Microsoft HoloLens 2, and it allows the user to view the processed data and control the model street light. It was found that the application performed as desired, displaying information such as voltage, current, carbon emission, electricity price, battery state of charge and LED mode, while enabling control over the model street light. Moreover, the Deep Q-Network machine learning algorithm successfully scheduled to buy electricity at times of low price and low carbon intensity, while the Long Short-Term Memory algorithm accurately forecasted traffic flow with mean RootMean-Square Error and Mean Absolute Percentage Error values of 12.0% and 20.0% respectively.