Consider the Accuracy Required, Scale of Application, and Nature of the Decision When Applying Artificial Intelligence and Machine Learning Solutions.
Implementation of Two Pilot Projects in Missouri Gave State Department of Transportation Staff Exposure to the Technology’s Opportunities and Constraints.
St. Louis, Missouri, United States
Data Acquisition and Processing Using Artificial Intelligence and Machine Learning
Summary Information
Recent trends in big data combined with technological breakthroughs in Artificial Intelligence (AI) and Machine Learning (ML) have the potential to transform transportation planning and operations. State departments of transportation are recognizing potential benefits, such as personnel time savings, data-informed decision-making, and enhanced worker safety. Thus, the Missouri Department of Transportation (MoDOT) evaluated opportunities offered by AI and ML to support agency operations. MoDOT tested two pilot projects where AI/ML could be applied: a highway median inventory and annual average daily traffic (AADT) factor grouping. This project took place between June 2022 and June 2024 at MoDOT’s St. Louis Transportation Management Center.
Key lessons learned include the following:
- Consider the accuracy required, scale of application, and nature of the decision when evaluating AI/ML solutions. MoDOT concluded that AI/ML was most likely to be cost effective when a desired decision is clear and quantitative, when robust training data is readily available, and when the resulting algorithm would be used at minimum 10,000 times.
- Involve Information Systems (the agency’s information technology department) early and throughout any AI/ML project to ensure compliance with department standards. This will help ensure that solutions are compliant with the existing technology stack (e.g., the programming language and operating system).
- Identify internal capacity to implement or develop bespoke AI/ML solutions. Alternatives include establishing new internal positions, training existing staff, establishing committees to review third-party solutions, and contracting out for services to address particular needs.
- Remember that the full potential of AI/ML may not be realized by substituting algorithms into existing processes. Rather, imaging entirely new processes and structures to take advantage of AI/ML capabilities could be required.
