Applying AI and machine learning to enhance transportation safety and mobility

How DOTs are leveraging advanced data analysis and machine learning algorithms to strengthen their transportation systems

By Rakesh Nune | HNTB

Departments of Transportation are investing in their physical and digital infrastructure to enhance safety, streamline mobility and inform long-term investment. These efforts are unfolding in a rapidly evolving landscape marked by the convergence of three major developments::

• Vast volumes of readily available data from sensors, traffic cameras, mobile apps, connected vehicles and crowdsourced platforms
• Advances in AI and machine learning algorithms capable of parsing and interpreting that data in real time
• Mounting operational requirements on agencies as transportation systems (roads, bridges, etc.) grow more complex

Across the country, transportation leaders are beginning to unlock the role technology can play in making roads safer and trips more predictable. Artificial intelligence and machine learning are being used to synthesize data and recommend timely, actionable interventions in the development of projects and in day-to-day operations. By automating complex tasks and enabling real-time decision-making, AI equips agencies to anticipate problems, intervene earlier and respond more effectively.

Sharper tools for safety and congestion management

Two of the most urgent challenges facing transportation agencies today — roadway safety and traffic congestion — also are the areas where AI offers some of the most
immediate benefits.

AI can support faster and more targeted safety improvements by helping agencies understand how people move through complex environments. By analyzing patterns from camera footage or near-miss incidents, AI can help reveal risks that might otherwise go unnoticed and inform smarter infrastructure and policy decisions. Rather than relying solely on historical crash data, agencies can leverage predictive algorithms to identify dangerous interactions between vehicles, cyclists and pedestrians before a crash occurs.

Similarly, AI enables more dynamic and predictive traffic operations. Instead of reacting to congestion after it has formed, AI tools can detect anomalies in real time, forecast traffic buildup and provide actionable information for DOT staff. Leveraging the predictive analysis from AI, agencies can adjust signal timing or reroute traffic to communicate with the traveling public, improving coordination across modes and corridors.

The following case studies show how DOTs are already using AI to improve congestion and safety.

Improving mobility and safety through AI-powered insights

Virginia Department of Transportation

The Virginia Department of Transportation (VDOT) is leveraging artificial intelligence to address two distinct but interconnected challenges: responding to unexpected traffic congestion and improving pedestrian safety. Through the Regional Multimodal Mobility Program (RM3P), VDOT is piloting AI models that forecast congestion up to 30 minutes in advance, identifying unusual traffic spikes rather than routine slowdowns. These forecasts trigger responses such as adjusting signal timings on alternate routes, deploying safety patrols, opening express lanes, alerting drivers via apps and signage, and testing incentive programs to shift travel patterns.

At the same time, VDOT is using AI-driven image recognition to develop a comprehensive inventory of sidewalks and crosswalks across its 80,000-mile roadway network. By analyzing satellite imagery, the system detects pedestrian infrastructure with 80–95% accuracy in urban areas, significantly reducing the time and cost of manual fieldwork. Together, these initiatives enable VDOT to enhance real-time traffic operations while also supporting long-term planning efforts to prioritize investments that improve accessibility, equity and multimodal safety.

Assessing safety conditions at bike and pedestrian conflict zones

District Department of Transportation

With streets in Washington D.C. increasingly shared by pedestrians, cyclists, scooters and vehicles, DDOT sought a better way to assess safety conditions. Instead of deploying new sensors, the agency used AI to analyze footage from existing traffic cameras.

AI tools track movement patterns and detect "near misses" — instances where people or vehicles came dangerously close without an actual collision. These insights provide a granular view of user behavior at intersections and along corridors, helping DDOT identify hot spots for safety improvements. The data is also being used to inform signal timing, signage and potential physical design changes.

Considerations for agency AI adoption

Laying the groundwork for responsible, scalable AI integration is essential to realizing its full potential in helping enhance transportation safety. The following considerations can support DOTs as they seek to introduce AI into their workflows.

• Identify needs and priorities. Begin by identifying strategic opportunities — such as improving crash response times, supporting asset inventories or mitigating congestion — and use those scenarios to drive your exploration of AI tools. The most successful applications are grounded in well-defined use cases that reflect operational priorities.
• Inventory your data. AI is only as effective as the data it draws from. Agencies can begin by conducting a thorough audit of available data streams: traffic sensors, camera feeds, satellite imagery, maintenance logs and even crowdsourced inputs from apps like Waze. Ensure datasets are accurate, current and structured for use. Strong data governance, including naming conventions, storage protocols and access permissions can help support transparency and replicability.
• Engage cross-functional stakeholders. AI projects often span multiple departments. Early collaboration between IT, planning, operations, safety, procurement, and legal teams helps identify constraints, anticipate integration needs, and foster organizational buy-in. This cross-functional approach also increases the likelihood that pilot programs evolve into enterprise-wide solutions.
• Invest in policy. As agencies implement AI, they must also develop guardrails. Policies should address data sourcing, privacy protections, transparency in algorithmic decision-making, model explainability, and ethical review processes.Agencies may also consider how to handle personally identifiable information (PII), third-party data partnerships, data retention timelines and requirements for a human in the loop.
• Pilot and iterate. Start with limited-scope implementations that can demonstrate tangible benefits. Early pilots allow agencies to refine methodologies, build internal comfort with the technology, and establish workflows for future scaling. Measure outcomes against baseline conditions to evaluate effectiveness and inform broader adoption.