Data Quality Assessment for Smart Corridor Digital Twin Development

Abstract: Smart corridor digital twins of are often created for the development and evaluation of emerging intelligent transportation systems and Connected and Autonomous Vehicle (CAV) technologies.
However, the development and implementation of smart corridor digital twins requires significant real time data. Currently, limited guidance exists for digital twin data quality assessment. To address this challenge, this paper presents a data quality assessment process applicable to the development of a real-time, data-driven, traffic microscopic simulation model, i.e., a smart corridor digital twin. The key steps in this process include: 1) data availability and requirement investigation, 2) data characteristic assessment, 3) data quality assessment, and 4) the evaluation of the impact of data imputations/errors on simulation results. Additionally, to support the digital twin development where data challenges are found the paper presents real-time data anomaly detection and imputation methods.
The presented processes are based on experienced gained from two previous digital twin development efforts undertaken by the authors: 1) the North Avenue Smart Corridor in Atlanta, Georgia, and 2) the Martin Luther King Smart Corridor in Chattanooga, Tennessee. For both smart corridors real-time traffic volume, turn movement counts, and Signal Phasing and Timing (SPaT) data streams were used to drive the digital twins. These two efforts are utilized to demonstrate the application of the data quality assessment, anomaly detection, and imputation methods.
Key findings from the data quality assessments of the two smart corridors revealed that for a real-time digital twin application development: 1) the presence of data gaps in real-time smart corridor data streams can have significant impacts on simulation performance measures, 2) it is crucial that the real-time SPaT data align with the modeled timing plans and signal phase layouts, 3) having consistent and predictable resolution of data streams is critical to the development of a reliable real-time digital twin architecture, and 4) understanding data record timestamp generation is a key step in accurate implementation of real-time digital twin applications.