1. Home
2. Ideas
3. Applying Artificial Intelligence and Mac...

Applying Artificial Intelligence and Machine Learning Advances for Freeway Operations

• Aug 4, 2020

Research Description

In the U.S.’s congested urban areas, freeway operators are increasingly challenged to keep their systems running efficiently while maintaining an acceptable level of service (LOS). From 2017 to 2019, the average annual time lost due to congestion by American drivers has increased by two hours as economic and urban growth continue nationally . This trend will persist in the long-term as demand for mobility and vehicular travel increases, leaving traffic management officials looking for ways to mitigate the adverse impacts resulting from overburdened freeway systems. The tools currently at the disposal of Traffic Management Centers (TMC) – reliance on human operators, embedded roadway sensors, information dissemination, and programmatic options, such as increasing capacity or adapting existing freeway capacity to flexible uses – are not all that is needed for increasingly complex and interwoven systems. To combat these issues, advanced, adaptable, scalable, and robust analytic tools are needed in order to uncover traffic pattern insights and optimize capacity where possible.

The emergence of Artificial Intelligence and Machine Learning (AI/ML) has the potential to provide freeway operators (e.g. TMCs) with new tools and opportunities to manage these critical assets, yielding safer, more resilient and more reliable transportation systems. In this statement, AI is referred to as the broad concept of enabling a device (e.g. computer) to perceive its environment and take appropriate actions to maximize the chances of achieving a specific goal. ML is usually considered as one of the study areas under AI, and focuses primarily on the ability of a device to learn through experience (e.g. historical data).

Research is needed to maximize the benefits from AI/ML technologies in response to the challenges identified above. In the immediate short-term, research is needed to develop a guidance of the best practices for the development and deployment of AI/ML in freeway operations. The guidance will assess relevant AI/ML deployment projects worldwide and review findings and lessons learned to recommend metrics and measures for freeway operators to track the progress and to achieve desirable outcomes at major milestones.

Additional Supporting Information

Benefits The major benefits for freeway operations gained from AI/ML tool deployment fall into two domains: 1. improved classification of objects, events, and patterns to enable faster and more informed decision making, and 2. increased accuracy in predicting trends/behaviors. Accurate prediction is critical in helping facility operators transition from a reactive to a proactive management strategy to minimize the negative impacts of different events on freeway operations.

Overall, AI/ML provides the opportunity to turn enormous amounts of complex, high-dimensional data into insightful, meaningful actions. Some of the ways traditional Transportation System Management (TSM) solutions could benefit from AI/ML include: - automatic detection and cataloguing of capital asset maintenance status; - automatic incident detection and classification; - automatic re-routing suggestions to travelers based on short-term traffic predictions; - automatic adjustment to the toll schemes to proactively preserve the capacity in managed lanes or toll roads; and - coordinated adjustments to the metering rates at on-ramps along a freeway corridor based on changes in travel pattern.

In addition, such AI/ML techniques can be scaled up to handle huge data streams (e.g. big data) in the foreseeable future that no human operators can. It is worth noting that the benefits gained from deploying AI/ML in freeway operations would be greatly enhanced if coupled with other emerging technologies such as Connected and Automated Vehicles (CAV), Internet of Things (IOT), remote sensing, and 5G. For instance, large amounts of higher-quality data could be generated by CAVs, in turn, reducing the need for freeway owners and operators to invest capital to construct and maintain static sensors. This CAV generated data would be transmitted to TMCs via vehicle-to-grid (V2G) and vehicle-to-everything (V2X) communications infrastructure. With the CAV data, AI/ML algorithms will be able to produce insights and action sooner and with higher precision and accuracy then currently attainable. TMCs would be able to adjust in infrastructure (e.g. variable speed limits, dynamic lane availability, etc.) or send route assignments to individual or groups of vehicles. The resulting network would be much more efficient and robust, capable of detecting and predicting disturbances to LOS and managing it with more tools and with greater fidelity.

---

Submitted By

Jim Katsafanas

Michael Baker International

412-269-4635

Attachments