The National Academies

NCHRP 07-34 [Anticipated]

Toward Artificial Intelligence-Enabled Decision Support Systems for TSMO Applications

  Project Data
Source: Louisiana Department of Transportation
Funds: $450,000
Staff Responsibility: Zuxuan Deng
Fiscal Year: 2023

This project has been tentatively selected and a project statement (request for proposals) is expected to be available on this website. The problem statement below will be the starting point for a panel of experts to develop the project statement.

As contemporary transportation systems get more complicated, it becomes much more challenging for decision makers to consider a large number of intertwined factors to optimize the systemwide processes, including planning, operation, asset management, and maintenance. For example, when an accident occurs, operational strategies need to be put forward to provide timely and effective multimodal services, such as vehicle routing, ramp metering, variable speed limits, emergency vehicle preemption, hard shoulder running, and adaptive traffic signal control to improve traffic incident management performance. Over the years, decision support systems (DSSs), which are primarily computer-based information systems used to sort, rank, or choose alternatives, have been developed to help infrastructure owners and operators (IOOs) and policy makers to gear transportation systems towards favorable directions. However, conventional DSSs are usually built on a set of expert rules that might not be able to provide customized and optimal solutions. On the other hand, artificial intelligence (AI), especially advanced machine learning (ML) technique, has been revolutionizing every facet of daily life, including transportation. It takes advantage of the availability of a massive amount of real-time data to model system behaviors, predict traffic states, and evaluate overall performance, which is well aligned with the key functions of DSSs. Therefore, there is a need to explore AI potential for transportation DSSs.

As an effective tool to support offline planning and online operation, transportation DSSs have received much attention from practitioners, researchers, and policy makers in the past few decades. They are widely used for land use planning, networked traffic assignment, logistics and supply chains, congestion or bottleneck mitigation, traffic incident management, and fleet/asset repair and maintenance. However, most of these DSS tools are rule-based or model-based without taking full advantage of available data from various sources. Recent advances in artificial intelligence, such as deep neural networks, have unlocked a myriad of opportunities to improve transportation systems, such as the development of connected and automated vehicles. Relatively few studies and deployments have been focused on the exploration of AI or machine learning to the development of data-driven DSSs for traffic system management and operations.

The objective of this research is to leverage the state-of-the-art development in artificial intelligence and machine learning, and explore their potential to improve DSSs mainly for transportation system management and operations (TSMO). The research should address (at a minimum) the following questions:

1. General: (a) What is the state of the art about decision support systems for TSMO? (b) What are the gaps of existing DSSs for TSMO? (c) What is the state of the practice for the application of artificial intelligence and machine learning in DSSs for TSMO?

2. Data: (a) What would be the minimum requirement about data (such as data sources, contents, spatial/temporal resolutions, and data quality) to enable AI application for TSMO? (b) Are there any innovative ways to collect data (while keeping privacy) necessary to support or improve AI-enabled DSSs for TSMO applications?

3. Methodology: (a) What are the suitable machine learning-based methodologies and tools for data pre-processing (e.g., cleaning, fusing) in DSSs for TSMO applications? (b) What are the appropriate AI tools that TSMO should use to solve specific functionalities?

4. Digital infrastructure: (a) What kind of sensors are suitable for different functionalities to collect data for AI-enabled TSMO applications? (b) What kind of communication means are required to support reliable information transmission for AI-enabled DSSs for TSMO applications? (c) What kind of computational power is needed to support timely decision-making in AI-enabled DSSs for TSMO applications?

Potential research tasks include:

1. Identify success stories and lessons learned from AI-enabled TSMO applications.

2. Conduct a survey on the basic requirements for a DSS in terms of spatial context, operator engagement, decision latency, proactive capabilities, cross-facility and cross-mode coordination.

3. Analyze the TSMO’s basic functionalities (e.g. demand prediction, travel time prediction, development of action plans and schemes, human-machine interaction, etc.) and identify the role of AI in facilitating these functionalities based on the latest state-of-the-art and state-of-practice.

4. Design a roadmap toward AI-enabled DSSs technology development and adoption in terms of data requirements, algorithmic aspects, validation and testing, computation needs, performance evaluation, infrastructure readiness levels, and workforce training requirements.

5. Demonstrate the developed roadmap through the selection of two use cases to evaluate readiness, identify infrastructure, technical, and workforce gaps, and develop strategies for closing these gaps.


To create a link to this page, use this URL: http://apps.trb.org/cmsfeed/TRBNetProjectDisplay.asp?ProjectID=5315