Kirk Steudle, national transportation advisor with Steudle Executive Group and former director of Michigan DOT, gives his insight into the rapidly evolving world of AI, explains how it has already changed traffic management, and predicts what to expect in the future
Last year, I explored the safety improvement opportunities derived from data sharing and more connected technologies. I was encouraged by the numerous sessions on new and existing programs from the 105th TRB Annual Meeting that were focused on applying AI for more accurate modeling, addressing persistent challenges. This brings me to my focus for this year.
It’s no secret that AI is poised to redefine many industries, including how we manage our road networks. It will do it by fundamentally changing how infrastructure owners and operators (IOOs) understand and respond to conditions across entire road networks. However, the advantage will be to the users who understand how AI is developed and deployed. This requires human intervention to grasp AI’s ability to synthesize realtime data with historical data and apply it as a tool for predicting congestion, forecasting conditions, while optimizing roadway operations.
“AI is able to infer not just that congestion is forming, but why”
Modern traffic systems can generate enormous volumes of data. This was good and bad just a few years ago. The good was that IOOs had data to verify operational performances. The bad was that this was a manual process and the rapidly growing amounts of data quickly became overwhelming and took away time from leveraging the technology effectively.
This is where AI-driven systems excel. They change the dynamic by reducing the operational burden of working across disparate datasets. Machine learning and analytics can synthesize realtime and historical data from multiple sources and formats to support forecasting demand surges, anticipating bottlenecks, and enabling proactive interventions. Generative AI builds on this by interpreting outputs, surfacing insights, and supporting decision-making at scale. But this requires human intervention and technical experience. Beyond numerical data streams, AI can also incorporate text-based information such as procedures, incident reports, and logs that historically sat outside traditional analytical workflows.
A core capability is data fusion. Instead of treating each data source independently, AI can blend heterogeneous inputs into a unified operating picture across different data formats. Fusing data inputs enables AI to infer not just that congestion is forming, but why.
30% – The reported reduction in waiting times at signals in Pittsburgh thanks to new AI analytics
Closely related is data integration. This process balances and matches information across disparate systems. Transportation networks rely on legacy infrastructure alongside emerging IoT devices that may or may not be focused on traffic management but can be a source of valuable data. AI-driven integration platforms standardize these inputs and eliminate time-consuming manual work. AI as a tool is ideal to sort through all the different datasets, but these systems still rely on human judgment. It ensures that analytics engines receive the relevant information suitable for high accuracy modeling and predictive algorithms, and that the results are applicable. I’m excited by the possibilities that AI is unlocking for the mobility ecosystem. Until next time, travel safe.
This aritcle was first published in the February/March edition of TTi magazine
