Digital Twins-Based Automated Pilot for Energy-Efficiency Assessment of Intelligent Transportation Infrastructure

• Published in: IEEE transactions on intelligent transportation systems 2022-11, Vol.23 (11), p.22320-22330
• Main Authors: Tu, Zhen, Qiao, Liang, Nowak, Robert, Lv, Haibin, Lv, Zhihan
• Format: Article
• Language: English
• Subjects: Analytical models; Automatic pilots; Data envelopment analysis; Data models; DEA; Digital twin; Digital twins; Efficiency; efficiency evaluation; Evaluation; Infrastructure; Intelligent transportation systems; Investment; LSTM; Prediction models; Predictions; Predictive models; Traffic flow; Traffic models; Transportation; Urban areas
• ISSN: 1524-9050; 1558-0016; 1558-0016
• DOI: 10.1109/TITS.2022.3166585

To realize the great potential of the intelligent transportation infrastructure, the investment in the transportation infrastructure in the intelligent transportation system should be rationally planned. Firstly, the application status of cutting-edge Data Envelopment Analysis (DEA) model in transportation infrastructure efficiency evaluation is analyzed, and based on this, a DEA model of transportation infrastructure efficiency evaluation under Digital Twins technology is established. Secondly, with the transportation infrastructure of 12 prefecture-level cities in Jiangsu Province from 2005 to 2020 as the research object, the Digital Twins DEA model and the traditional Stochastic Frontier Approach (SFA) model are used to estimate the efficiency of transportation infrastructure in 12 cities. Finally, the traffic flow data of a certain road section in Zhenjiang City (J11 City) is simulated and predicted by using the Long Short-term Memory (LSTM) traffic flow prediction model. The results show that the average efficiency of the 12 cities estimated by the DEA model based on the Digital Twins is 0.7083, the average efficiency of the 12 cities estimated by the SFA model is 0.6445, and there are significant differences in the efficiency rankings of the cities. Compared with the actual efficiency, the established Digital Twins DEA model is more reasonable for the calculation of transportation infrastructure efficiency. The results of the LSTM traffic flow prediction model show that the Mean Absolute Error (MAE) of the LSTM model is 24.29, the Root Mean Square Error (RSME) is 0.1186, and the Mean Absolute Perce (MAPE) is 17.78, which are all lower than other models. Compared with other models, the proposed LSTM-based traffic flow prediction model is more accurate in traffic flow prediction. Hence, the research content provides a reference for the investment planning of intelligent transportation system infrastructure.