DOT pilots AI SMART tool to cut flight delays

The U.S. Department of Transportation has begun piloting a new artificial intelligence supported air traffic management tool in the Washington region, seeking to spot congestion earlier in the day and reduce the flight delays and cancellations that routinely disrupt U.S. airline travel.

AI comes to air traffic management

The pilot centers on a system known as SMART, short for Strategic Management of Airspace, Routes and Trajectories. Publicly available material from the Department of Transportation and the Federal Aviation Administration describes SMART as a planning platform that pulls together more than 200 separate data streams, including airline schedules, weather forecasts, airport capacity, airspace restrictions and controller staffing metrics.

The software uses artificial intelligence techniques to synthesize that information into a single view of expected traffic flows. It then runs predictive models to identify where demand is likely to exceed capacity, where storms or turbulence could close off key routes and how those disruptions might ripple through the wider network over the course of a day.

Reports indicate that the tool is designed to work as an advisory system layered on top of existing air traffic infrastructure such as the En Route Automation Modernization platform and the FAA’s broader NextGen modernization program. Human controllers and traffic managers remain responsible for separation and safety, while SMART focuses on suggesting earlier, more efficient ways to reroute traffic before problems escalate.

DOT documents on artificial intelligence strategy frame this type of application as an example of using data driven prediction to improve system reliability without changing underlying safety rules. The agency has been cataloging its AI use cases across modes of transportation, emphasizing that any operational deployment must be explainable, auditable and compatible with existing certification processes.

Washington region airports serve as test bed

The initial SMART rollout is concentrated in the busy airspace around the U.S. capital. According to published coverage, the pilot involves Ronald Reagan Washington National Airport, Washington Dulles International Airport and Baltimore Washington International Thurgood Marshall Airport, three hubs that together handle thousands of flights per day across domestic and international networks.

These airports sit at the heart of a corridor that is prone to delays when summer thunderstorms pop up along the East Coast, when winter storms affect the Mid Atlantic or when traffic along key routes between the Northeast and Southeast becomes saturated. Historical performance data for the region has repeatedly shown that disruptions in this corridor can quickly ripple to other parts of the national system.

By running SMART in this highly constrained environment, the Transportation Department and FAA can test how well the system anticipates those bottlenecks and whether earlier route adjustments help keep departures and arrivals moving. Published descriptions indicate that a centralized SMART center at DOT headquarters synthesizes the predictions and shares recommendations with the existing Air Traffic Control System Command Center for possible action.

Early phases of the pilot are focused on validating the models and ensuring that the AI generated recommendations match real world conditions closely enough to be useful for daily operations. Observers will be watching for data on concrete outcomes such as changes in ground delay programs, reductions in holding patterns or improvements in on time departure rates during adverse weather events.

From reactive to predictive flight planning

For travelers, the significance of SMART lies less in the technical architecture and more in how it could change the timing of decisions that drive delays. Traditional air traffic flow management often reacts to storms or congestion after they begin to affect operations, which can lead to a cascade of ground stops, diversions and cancellations as the day progresses.

By contrast, the new AI powered platform is intended to support what aviation planners call trajectory based operations, a NextGen concept in which all stakeholders share a common understanding of planned flight paths in three dimensions plus time. The more precisely those trajectories can be predicted and updated, the easier it becomes to adjust departure slots, reroute flights around constrained airspace and balance demand across multiple airports before queues build up.

SMART’s predictive capabilities depend on continuously updated inputs. Weather models feed in forecasts of storm development along key jet routes, while airlines supply schedule and fleet data. Airport operators contribute information on runway availability and surface constraints, and the FAA adds information about temporary flight restrictions or sector capacity limits. The AI engine looks for conflict points in that combined picture and evaluates alternative routings that still fit within safety and capacity rules.

Public explanations of the program emphasize that any changes ultimately flow through established FAA processes. Recommendations from SMART are intended to complement, not override, controller judgment, which remains the primary safeguard for maintaining separation standards and responding to unexpected events in real time.

Balancing efficiency, safety and public trust

The emergence of AI in air traffic management comes at a time of broader public debate over automation in safety critical domains. Transportation research sponsored by DOT has highlighted both the potential benefits of AI and the need for robust safeguards, particularly in areas where algorithms interact with human operators responsible for life critical decisions.

Recent technical reports on aviation AI stress the importance of validation, uncertainty quantification and so called run time assurance, which refers to mechanisms that prevent an automated system from taking actions outside predefined safety envelopes. In the context of SMART, those principles translate into tight limits on what the software is allowed to recommend, clear audit trails for how suggestions were generated and the ability for traffic managers to accept, modify or reject those suggestions.

DOT’s consolidated inventory of AI use cases underscores a broader federal requirement to ensure that artificial intelligence systems used by agencies are transparent and accountable. The Advancing American AI Act and related guidance require agencies to document how these systems function, how data is used and what steps are taken to mitigate bias or unintended consequences.

For travelers, these governance frameworks are largely invisible, but they shape how quickly tools like SMART can scale from pilots to nationwide deployment. The more clearly the agencies can demonstrate that AI is supporting human expertise rather than replacing it, the easier it may be to build confidence among passengers, controllers and airline operators.

What travelers should watch next

Even in its early pilot phase, the SMART program has practical implications for frequent flyers and airport managers. If the system performs as intended, one of the first visible changes may be a reduction in the kind of cascading delays that strand passengers far from the original weather or congestion hotspot that triggered the disruption.

Travelers using the Washington area airports over the coming months may still encounter routine schedule changes, particularly during peak travel periods and severe weather. The key question is whether those adjustments become more targeted and less disruptive, with fewer last minute cancellations and a greater emphasis on rerouting and modest departure time shifts instead of outright ground stops.

Industry coverage suggests that results from the Washington pilot will inform decisions about expanding SMART to other congested corridors, such as the New York and Chicago regions. Any broader rollout would intersect with ongoing modernization programs like NextGen and updated flow management services that already aim to improve predictability and throughput in the national airspace system.

For now, the pilot marks one of the most visible attempts to apply artificial intelligence to a problem that travelers experience daily: how to keep complex, weather sensitive airline networks running on time. The data that emerges from this experiment will determine whether AI powered planning tools become a core part of how flights are managed across the United States.