Establish Stable Funding and Contracting Mechanisms to Maintain Specialized Software Development Services.
Delaware DOT Implemented and Evaluated an AI-Enhanced Integrated Transportation Management System Statewide.
Delaware, United States
Advanced Transportation and Congestion Management Technologies Deployment (ATCMTD): Artificial Intelligence-Enhanced Integrated Transportation Management System (AI-ITMS)
Summary Information
Transportation agencies acknowledge that the extensive volume of traffic data can be a challenge to manually process for continuous monitoring, assessment, and efficient response. In this study, Delaware DOT (DelDOT) harnessed the power of technological advances in Artificial Intelligence (AI) and Machine Learning (ML) to create an AI-enhanced Integrated Transportation Management System (AI-ITMS). This software platform, when deployed, would potentially improve how the DelDOT Transportation Management Center (TMC) monitors the transportation system by predicting traffic flows; identifying anomalies and inefficiencies; and generating solutions to existing and predicted traffic congestion. This study provided a status update for the AI-ITMS, as well as lessons learned from the overall system development process.
- Establish stable funding and contracting mechanisms to maintain specialized software development services. This mechanism should be continuously evolving and updated to keep up with the advances in technology, especially in the areas of data science and algorithm development.
- Consider data quality monitoring to enhance and ensure AI and ML model usefulness. Robust data quality monitoring strategies, including redundancy mechanisms and backup data sources, should be implemented to ensure continued model functionality, even when primary data is corrupted or missing. Using training models with tolerance for incomplete or erroneous data is recommended to help maintain their effectiveness in real-world scenarios.
- Bring design standardization to simplify the software design efforts. This would be especially needed in detection layout and associated cabinet wiring standards.
- Be aware of the challenges associated with deploying AI as a tool. This study pointed out efforts related to AI in transportation and the need for establishing processes for applying traditional project tasks (e.g., data collection, deployment, and evaluation techniques) to the non-traditional AI program.
- Evaluate Long-Term Computing Architecture for Cost-Effective and Scalable Performance. Future system scalability necessitates assessing the balance between on-premises and cloud-enabled capabilities. Consideration of AI/ML, modeling, and simulation data storage and computing needs will guide cost-effective infrastructure choices.
- Establish a true baseline for system evaluation to enable comprehensive benefit-cost analysis. A well-defined baseline should assess quantifiable performance measures and track if the expected time and cost goals were met. This study suggests starting with a small number of affected segments can help refine delay impact variables for accurate system benefits assessment.
