NCHRP 23-12

ARTIFICIAL INTELLIGENCE OPPORTUNITIES FOR STATE AND LOCAL DOTS – A RESEARCH ROADMAP

Appendix E: Research Problem Statements

Prepared for NCHRP
Transportation Research Board
of
The National Academies of Sciences, Engineering, and Medicine

Aditi Manke
Abhijit Sarkar
Matthew Camden
Alejandra Medina
Tammy Trimble
Christie Ridgeway

Virginia Tech Transportation Institute
Blacksburg, VA

Introduction

Based on the results from the previous outreach efforts of interviews and workshops as well as the feedback received from the panel members, the research team finalized 11 research problem statements. The research team also referenced the National Artificial Intelligence R&D Strategic Plan published in 2019. The following eight key strategies highlighted in that report were considered during the development of the problem statements:

1. Make long-term investments in AI research.
2. Develop effective methods for human-AI collaboration.
3. Understand and address the ethical, legal, and societal implications of AI.
4. Ensure the safety and security of AI systems.
5. Develop shared public data sets and environments for AI training and testing.
6. Measure and evaluate AI technologies through standards and benchmarks.
7. Better understand the national AI research and development workforce needs.
8. Expand public-private partnerships to accelerate advances in AI.

The problem statements are discussed in detail with background information on why this research needs to be conducted and the expected tasks, budget, and duration necessary to achieve required research objectives. The research team identified six research areas where they felt the proposed draft roadmap ideas would address the problems in those areas. The six research areas are:

1. Workforce and infrastructure development.
2. Readiness and evaluation of AI.
3. Challenges in adopting AI.
4. Current practices and prioritization.
5. External collaboration.

Equity, policy, and planning Table 24 shows how the draft problem statements covers one or more of those research areas.

Table 24. Grouping the roadmap ideas by research focus areas.

Research Problem Statement 1: Case Studies of Implementation of Artificial Intelligence Programs in State and Local Departments of Transportation

Background

AI is increasingly available for state and local DOTs to solve their transportation challenges. However, most applications of AI within the transportation industry are still from early adopters of the technologies. To further accelerate the adoption of AI within state DOTs, results from NCHRP Project 23-12 Artificial Intelligence Opportunities for State and Local DOTs – A Research Roadmap show that state and local DOTs need independent evidence on the application and effectiveness of AI programs. This finding highlights the need for additional research to identify challenges, benefits, opportunities, and effectiveness data on the use and adoption of AI technologies from early DOT adopters.

Potential Benefits

This proposed activity will help develop case studies and document lessons learned from the early adoption of AI within DOTs. The case study results will gather critical data on challenges faced by DOTs when implementing AI, strategies to overcome barriers, cost data on AI deployment AI, workforce development issues, and effectiveness data. These data can be used by other DOTs in considering, planning, and implementing AI within their agency.

Objective

The objective of this research will be to document case studies of the successful implementation of AI programs within state and local DOTs to improve the efficiency or safety of the transportation system.

Tasks

• Conduct a literature review to identify effective AI solutions within state and local DOTs.
• Survey state DOTs to identify applications of AI and for agencies willing to participate in in-depth case studies on the use of AI to improve efficiency and safety.
• Identify metrics to define successful studies that uses AI.
• Develop case studies of state and local DOTs that have used AI in transportation-related applications. Each case study will collect qualitative and quantitative data on the application of AI, including the functional area where AI was deployed, challenges faced in AI implementation, strategies to overcome challenges, costs of implementing AI solutions, sources of funding that supported the deployment of AI, policy considerations for the implementation of AI, and effectiveness and benefits of AI technologies.
• Conduct a series of webinars and conference presentations to share case study results with state and local DOTs to accelerate the adoption of effective AI solutions.
• Prepare a final technical memorandum documenting the research process and results.

Expected Deliverables

• Final Report outlining the methodology, results, and conclusions from each phase of the project.
• Research dissemination plan.

Estimated Funding:

$250,000

Expected Duration:

18 months

Research Problem Statement 2: Toolbox to Guide the Selection and Deployment of Artificial Intelligence Technologies in State and Local Transportation Agencies

Background

AI, including Machine Learning, Natural Language Processing, computer vision, big data analysis, deep neural networks, and multimodal sensor processing, has experienced unprecedent advancements in the previous 10 years. These advancements improved the ability to process large-scale data and high-speed computing to achieve faster and more accurate results. With the growth of these technologies, many enterprises and researchers have developed useful AI-based solutions that can be used to solve challenges in diverse applications, including transportation. For example, state DOTs and local transportation agencies have utilized AI-based programs for traffic incident detection, traffic flow analysis, identification of pedestrian and other vulnerable road user traffic patterns, evaluation of roadway conditions, evaluation and planning of winter maintenance activities, and improving access to transportation-related information, among many other areas.

Although the potential applications of AI-based solutions increased in recent years, transportation agencies often lack the information and guidelines for the selection, prioritization, and deployment of AI-solutions. As documented in NCHRP Project 23-12 Artificial Intelligence Opportunities for State and Local DOTs – A Research Roadmap, state DOTs and local transportation agencies identified a lack of information, tools, and educational resources as major barriers for the implementation of AI within their agencies. As reviewed in this report, transportation agencies need guidelines and tools to prioritize effective AI-based solutions and to accelerate their adoption to realize the benefits. This informational toolbox should be designed to help transportation agencies identify the following:

• Readiness for and Potential applications of AI technologies in state and local transportation agencies.
• AI-based technologies with potential to address DOTs challenges.
• The costs associated with AI solutions.
• The potential benefits of AI solutions.
• The potential return-on-investment of AI solutions.
• Characteristics and parameters to consider in identifying a solution to deploy.
• Communication strategies to educate business functions within the agency about potential AI solutions.

Potential Benefits

This toolbox will facilitate decisions by transportation agencies in determining where to implement new AI technologies within DOTs and which technologies are likely to be the most beneficial. The tools and information developed in this project will provide transportation agencies with evidence-based, objective data to inform decisions on AI solutions and to offer parameters and guidelines to improve the implementation of effective technologies.

Objective

The primary objective of this project is to develop decision tools and guidelines transportation agencies can use in assessing and deploying effective AI solutions. It is expected that this toolkit will help agencies to evaluate the readiness of AI technologies and prioritize the deployment of AI projects.

Tasks

• Conduct a national and international literature review of (a) transportation AI technologies, and (b) the criteria used to identify business areas for AI investment, evaluate readiness, and guide the implementation of new technologies in transportation agencies.
• Conduct a survey of state DOTs of current practices regarding the selection and prioritization of AI mechanisms to improve performance measures.
• Prepare a list of available and emerging AI solutions for transportation agencies and organize them in a toolbox that allows assessment of a business function’s readiness to implement AI (data quality, workforce capabilities, process clarity, etc.), prioritize AI investments, and evaluate potential challenges and opportunities associated with each technology.
• Develop a series of one-page educational materials that transportation agencies can use to increase leaderships’ knowledge of AI.
• Prepare a final technical memorandum documenting the research process and findings.

Expected Deliverables

• An electronic toolkit (with links to supporting documentation) that provides resources to identify business areas for which AI may be beneficial, assess AI readiness within an organization, evaluate workforce capabilities for AI development and deployment, evaluate policy needs, as well as information on the common challenges and potential solutions, and sources of funding that may be used for the development and deployment of AI.
• A series of one-page educational materials that transportation agencies can use to increase leaderships’ knowledge of AI.
• A final report documenting all information gathered, methodologies, and results.
• A list of available and emerging AI solutions within the electronic toolkit with information to help decision makers understand potential applications and challenges with each technology.
• Implementation guidelines to assist decision makers in identifying, prioritizing, and deploying AI solutions.

Estimated Funding:

$300,000

Estimated Duration:

24 months

Research Problem Statement 3: Workforce Needs and Development to Prepare Transportation Agencies for the Application of Existing and Emerging Artificial Intelligence Approaches

Background

Implementing changes to accelerate the adoption of Artificial Intelligence (AI) approaches for transportation-related applications relies on educating the workforce in general, and DOT personnel specifically. Some states have already acknowledged workforce requirements to meet the needs of incorporating AI technologies in everyday and future processes. States have identified problems relating to the availability of a capable workforce, the definition of the skills that will be needed, and competition with private industry for qualified workers. Further, those who do have a strong background in AI do not generally have experience with transportation applications. In thinking about the application of AI technologies to create better solutions to transportation problems, agencies mention an expectation from upper management for the agency as a whole to support the application of AI technologies. However, they tend to underestimate the need for resources in general, and the workforce in particular. This lack of workforce expertise within DOTs related to machine learning restricts states’ abilities to promote AI projects.

While state DOTs agree there are benefits to promoting and finding trustworthy partners in universities and private industry, there is also agreement on the need for a strong state DOT counterpart workforce throughout the process. Strategies to address these workforce challenges have included hiring employees with the necessary capabilities, working with partner agencies and other stakeholders (e.g., contractors, industry partners, and the general population); establishing working groups focused on workforce development activities; and supporting employees as they pursue the acquisition of new skills. To maximize the benefits associated with the identification and implementation of transportation-related AI projects, there is a need to identify the workforce skills, education, and training for the development and implementation of AI solutions. In the near term, this could include identifying the necessary skills, developing skills courses, and making the courses available to DOT personnel. In addition, DOTs should encourage employee participation at national and local AI-related forums and encourage peer-to-peer knowledge transfer within and between DOTs.

Potential Benefits

By working to identify the necessary skills to implement AI in transportation-related solutions and the necessary resources to develop the associated skills and training, DOTs will be better prepared to develop and implement AI solutions moving forward.

Objective

The objectives of this research are (a) to identify workforce AI development, deployment, and management needs, for and (b) to gather and/or develop job descriptions for AI-related positions, and (c) to provide recommendations for developing and deploying the required training/certifications. The research must identify current workforce needs and the associated strategies for building capacity into the future as technology evolves.

Tasks

• Conduct a literature review regarding the application of AI in transportation-related solutions and the associated workforce needs. Please note that the workforce needs should include the skills and

• education needed by supporting personnel (e.g., data collection and management, cybersecurity) as well.
• Survey stakeholders regarding the skills and training needed by DOTs to support the application of AI transportation-related solutions. Stakeholders should include state DOTs and relevant agencies, private industry, academia, and associations such as the American Society of Civil Engineers and the Institute of Transportation Engineers.
• Based on the previous two tasks, collate the knowledge, skills, and abilities needed to support AI. This effort should include:
• The identification of the steps to be taken to develop the workforce. Particular attention should be given to gaps in the industry; between college graduates and the work needed, and what it will take to fill that gap. Training may include in-house efforts, university and/or private industry partnerships, and should encompass personnel with AI backgrounds and personnel supporting AI efforts such as data collection and processing.
• The identification of the skills and associated training needed to support different levels of AI implementation. As part of this effort, researchers should gather and/or prepare position descriptions.
• The outcome of this task should be a list of new or modified positions needed, corresponding skills, and training mechanisms. The product should be incorporated into the Agency Capacity Building (ACB) portal (https://www.agencycapability.com/)
• Present and discuss findings with stakeholders in focus groups and/or workshops to refine the recommendations.
• Prepare a final technical memorandum documenting the research process and findings, add draft job descriptions to the ACB portal, develop outreach materials.

Related Work

• NCHRP 20-102(20) Preparing Transportation Workforce for the Deployment of Emerging Technologies
• AASHTO Joint Task Force on Digitalization

Expected Deliverables

• Final report defining the skills and training needed for different worker categories.
• List of available resources including training and academic programs.

Estimated Funding

$250,000

Expected Duration

24 months

Research Problem Statement 4: Implementable Funding Strategies for Artificial Intelligence Opportunity Applications for State and Local DOTs

Background

The U.S. DOT considers AI to be one of the instruments with the potential to advance transportation through revolutionary solutions to transportation challenges. Based on the outreach conducted for NCHRP Project 23-12 Artificial Intelligence Opportunities for State and Local DOTs – A Research Roadmap, the lack of available funding for AI initiatives was seen as a major barrier to implementation. It was noted that funding opportunities call for projects that support the integration of new technologies; however, very few of these opportunities specifically reference AI. Additionally, DOTs have limited resources and budgets for AI. As a result, the lack of funding information and limited resources lead to hesitation to propose the use of AI in existing or future transportation solutions. Thus, there is a need to identify how state and local DOTs can use existing and new funding mechanisms to test and incorporate AI into transportation processes.

Potential Benefits

This proposed activity will help raise state and local agency awareness of the different avenues available for obtaining federal and regional funding to incorporate AI in transportation agency projects. The results of this research may also provide smaller local agencies with the information they need to better compete for funding.

Objective

The objective of this research will be to identify existing and new funding mechanisms for the development, testing and incorporation of AI into existing and future transportation processes. Additionally, this research will characterize the best practices associated with the estimation of project costs and the identification of matching funds.

Tasks

• Inventory current multimodal federal and regional funding mechanisms to identify those that have potential for use in the development and implementation of AI in transportation-related applications.
• Develop case studies of state and local DOTs that have successfully used federal and regional funding mechanisms for the implementation of AI in transportation-related applications. Case studies should include the criteria used to prioritize the funding sources and to determine success.
• Drawing upon the findings from the previous two tasks, draft a list of possible funding sources and the challenges and opportunities for using these resources to advance AI. Review the findings with a focus group of agency leadership and budget program managers. Revise materials as appropriate to incorporate focus group feedback.
• Prepare outreach materials for agency leadership and senior managers. Hold a series of webinars highlighting specific funding mechanisms that may be leveraged.
• Prepare a final technical memorandum documenting the research process and results. Prepare outreach materials for agency leadership and senior managers.

Expected Deliverables

• Final Report outlining the methodology, results, and conclusions from each phase of the project.

• Research dissemination plan.

Estimated Funding

$250,000

Expected Duration

12 months.

Research Problem Statement 5: Develop a Guidebook to Understand the Vulnerability and Security Concerns for AI-based Transportation Solutions

Background

AI-based transportation solutions are often governed by black box models. Black box models refer to models. Modern AI-based methods provide high performance, but to the detriment of explainability. AI methods can also result in unwanted biases, depending on how they were developed. Adversarial attacks can make AI methods produce incorrect results and create security vulnerabilities. Lastly, most AI methods perform well in a specified domain, but fail to generalize in other domains. These limitations can make AI-based methods vulnerable in many ways. Therefore, a deeper understanding of AI methods and limitations is critical. This project aims to create a guidebook to help decision makers within transportation agencies better understand these possible limitations, biases, and vulnerabilities. These can vary by their application within a DOT, the complexity of the application, and the methods and toolsets used for the targeted solution. This project will highlight the risk of these limitations for various applications and create a guidebook for an explainability and testing regime that will promote efficient AI deployment.

Potential Benefits

• This research will identify potential threats and possible countermeasures in the implementation of AI transportation solutions.
• This research will help identify the sources of biases and possible steps to mitigate them.
• This research will help identify limitations of any solution to its application domain. This will guide DOTs to choose a particular AI-based solution depending on their needs, available resources (data, software, IT infrastructure), and expected outcome.
• This research will help broaden understanding of the scope of a particular AI solution, hence creating a guideline of dos and don’ts while using AI solutions.
• This effort will support upgrades to the IT infrastructure to protect from future attacks and threats.

Objective

This project will highlight the risks and limitations for various applications and create a guidebook for an explainability and testing regime that will promote efficient AI deployment. Explainable AI (XAI) is a growing field of research that offers explanation to many of the black box models. One additional objective of this research is to determine how XAI can be used for transportation research to guarantee robust solutions.

Tasks

• Identify and list modes of vulnerabilities for AI-based transportation applications or solutions. This should include potential solution biases (e.g., privacy, equity, cybersecurity, ethics, and legal concerns) and sources of adversarial attacks.
• For each application or solution identified in Task 1, provide an understanding of the criticality of failures. Through discussions with stakeholders (academia, industry, DOT practitioners and others as appropriate), rank the criticality of the associated failures.
• Perform survey of evaluation methods, criteria of the failure in different kinds of AI-based applications, and methods for assessing the robustness of methods. In this case, robustness may refer to the performance of the system under adversarial attacks.

• Create a guideline for testing and validation of different modes of AI applications and different failure modes associated with each application.
• Create a guidebook providing an overview of the data standards, infrastructure, and research needs necessary to understand vulnerabilities and security concerns.
  • Conduct case studies to demonstrate vulnerabilities and potential solutions.
  • Select a set of case studies by consulting the stakeholders from regulatory agencies.
  • Create a threat model for each problem.
  • Identify possible countermeasures.

Expected Deliverables

• A detailed report summarizing the possible sources of threat, their criticalities, and possible countermeasures.
• A robustness analysis plan and guidebook to DOTs for implementation of AI-based solutions.
• A report of case studies conducted on some potential application areas in transportation research.

Suggested Reference

• NCHRP Report 1034 Guidelines on Collaboration and Information Security for State DOTs
• Phillips, P. J., Hahn, C. A., Fontana, P. C., Broniatowski, D. A., & Przybocki, M. A. (2020). Four principles of explainable artificial intelligence. Gaithersburg, Maryland, 18.

Estimated Funding

$300,000

Expected Duration

24 months.

Research Problem Statement 6: Exploring the Integration of AI-based Methods in Multimodal Transportation Planning

Background

For a transportation system to be efficient and fair, it must serve diverse demands. For most of the last century, transport planning has heavily centered around automobiles. As a result, most communities have well-developed road systems that allow people to drive to most destinations. This kind of planning has left out non-automobile travel demands, and over the years the automobile-focused road system has seen an increase in travel time and traffic congestion.

Multimodal transportation planning is a comprehensive approach to urban or regional transportation systems that aims to integrate and optimize various modes of transportation to improve efficiency, accessibility, and sustainability. The goal of multimodal transportation planning is to create a seamless and convenient travel experience for users, allowing them to easily switch between different modes of transportation during their journeys. It promotes diverse modes of transportation, focuses on public benefits, and strategizes transportation planning based on the socioeconomic condition of a certain area, their land usage, and available connectivity between transportation hubs. As a result, this can facilitate addressing environmental sustainability, benefiting a large population of users, and promoting equity, especially for underprivileged groups.

Potential Benefits

The proposed project will educate city planners and transit agencies about different AI-based predictive models and how integrating these models into transportation management will benefit them in improving traffic monitoring and forecasting. Creating models that consider all modes of transportation will help planners in improving and redesigning transport systems that meet various user needs.

Objective

The objective of this project is to study some of the predictive models that look at reducing travel time and peak period congestion, determine some of the gaps and limitations in the existing models, and identify if new variables need to be considered in the predictive models. The project should also focus on data analysis techniques that model the travel demands as well as data on safety and safe travel perceptions of vulnerable road users.

Tasks

• Conduct a literature review highlighting AI and ML applications in multimodal transport planning.
• Document the current state of practices in understanding safety, traffic congestion, monitoring, and travel times in multimodal planning models. Identify planning models that include active transportation along with other modes of transportation.
• Highlight the gaps and limitations in the current predictive models.
• Develop a new predictive model that considers all the necessary data points and perform preliminary data analysis that predicts travel demands of various users. Explain how planners and transit agencies can use these results to redesign urban areas, improve transportation management, and increase travel satisfaction of users.
• Prepare a final technical memorandum documenting the research process and results.

Expected Deliverables

• A report on various AI-based methods that are available for transit agencies to use for transportation planning and management.
• A report that highlights the gap and data needs for current predictive models.
• A report on a new predictive model, documentation of data analysis, and results.

Expected Budget

$200,000

Expected Duration

24 months

Research Problem Statement 7: Validation of Artificial Intelligence Applications for Automated Pavement Condition Evaluation

Background

Advanced laser- and image-based pavement systems have been widely adopted by highway agencies in the last decade for automated pavement condition assessment, while 2D imaging technologies, smartphones, and tablets are also used to perform pavement condition evaluation, especially for local transportation agencies. The data collected are then used to extract pavement distresses automatically through various methods, often using AI. However, there are still pending challenges associated with the assessment of the accuracy and precision of the reported distress identification and measurements required for network- and/or project-level pavement management decisions.

An ongoing NCHRP synthesis is documenting state DOT current practices of automated pavement distress identification using AI (ML/deep learning) technologies for pavement condition evaluation. This effort will document the requirements for automated pavement distress identification; various applications of pavement distress condition information; types of agency decision-making supported by pavement condition data; AI technologies, tools, and models currently being used; and reference/benchmark data used in AI-technique development, training, and evaluation.

Potential Benefits

This proposed activity will help establish the required protocols and standards to facilitate the adoption of AI for pavement assessment within DOTs. This will help harmonize pavement condition evaluation nationwide and reduce the cost of pavement condition data collection.

Objective

The proposed research project will build on the findings of the synthesis to define processes, protocols, and baseline reference data sets to test and validate approaches and tools for automatic identification and quantification of pavement distresses. The outcome will be a series of proposed American Association of State Highway and Transportation Officials (AASHTO) standard practices and protocols to assess and validate automated pavement condition approaches, processes, and tools.

Tasks

• Conduct an expanded literature review on current approaches, preprocesses, and tools in use that apply to AI for pavement distress identification.
• Investigate other available technologies or approaches used in other fields or applications (e.g., medicine, cybersecurity) that have not yet been used for pavement condition assessment.
• Identify potential sources of data that can be used to compile/build a reference data set for different types of pavements and distresses.
• Develop draft standard practices and protocols to assess and validate automated pavement condition approaches, processes, and tools.
• Test the standards in collaboration with at least two state DOTs.
• Refine the methods as needed and develop the proposed AASHTO standard practices and protocols to assess and validate automated pavement condition approaches, processes, and tools.
• Prepare a final technical report documenting the research process, results, and implementation plan.

Estimated Funding

$500.000

Expected Duration

36 months.

Research Problem Statement 8: Explore Natural Language Processing-based Methods for Document Management and Public Interaction at DOTs

Background

In recent years, NLP has become immensely successful. NLP helps to automatically study texts and documents. This shows promise in minimizing manual effort in document management, text summarization, and customer service. Large language models (LLMs) like ChatGPT have shown potential to automate tasks like customer service and can provide quick automated responses and custom messages, minimizing human interaction time. DOTs handle a large volume of documents every day. This may include project reports, environmental assessments, traffic studies, contracts and agreements, budget and financial reports, and employee information. NLP can significantly help in maintaining and interpreting these documents while reducing human hours. The goal of this project will be to identify key areas and tasks at DOTs where NLP can be useful. The project will also identify a list of available NLP tools. Finally, the project will demonstrate the benefits of NLP in several example use cases.

Potential Benefits

• Integration of LLM can help DOTs to efficiently process text-based data.
• LLMs and chatbot models can help address customers (e.g., handling enquiries, resolving complaints, emergency responses, regulatory compliance, automated issue resolution).
• LLMs can help automate tasks like promoting safety messages, public relations, regulatory compliance, feedback documentation, and summarizing.
• LLMs can be used to automate customer feedback recording and summarization.
• Use of LLMs can facilitate 24/7 services, deliver quick and consistent replies, reduce wait times, and handle repetitive tasks, ultimately saving human hours and eliminating human errors.

Objective

• The objective of this research is to develop a guide, including implementation roadmaps, to help state DOTs and other transportation agencies in developing and deploying next-generation NLP-enabled systems.
• A key emphasis should be identifying the scope of recent tools that use LLMs, including services like ChatGPT, Bard, and Co-pilot, that can be implemented in DOTs and other transportation agencies.
• Two major areas of emphasis should be the use cases related to document management and public interaction.
• This research will also study and identify the benefits and limitations of these methods in specified use cases using testing and demonstration.

Tasks

Phase I: Planning

• Conduct a detailed literature review on the current state of NLP. The review should also identify and list opensource and commercial tools that are currently available. The review should identify the capabilities of each of these systems and perform an analysis of their merits and limitations.
  • Conduct a workshop to present the outcome of the literature review and discuss the strengths, limitations, and applications of NLP.

Phase II: Execution

• For each application, the following considerations should be reported and documented. To evaluate the efficacy of LLMs for DOT usage, it is important to study each.
  • Data requirements
  • Infrastructure requirements including computational resources.
  • Performance evaluation criteria
  • Operational challenges
  • Privacy concerns
  • Workforce requirements
  • Any additional considerations
• Conduct case studies with one or more agencies for LLM-enabled applications that were already identified in previous tasks.
• The outcome should be properly evaluated in terms of cost benefit analysis, performance, privacy protection, explainability, regulatory compliances, adversarial attacks, bias, and fairness.

Phase III: Final Products

• The outcome of all the tasks should be summarized as several technical memoranda and reports.
• The outcome should be disseminated through webinars or workshops.

Expected Budget

$550,000

Expected Duration²⁷

30 months

___________________

²⁷ This project can be divided into two projects depending on the application area focus.

Research Problem Statement 9: Develop a Guidebook for Successful Collaboration with Industry Partners that Provides AI-based Solutions.

Background

Currently, private companies lead in providing AI-based solutions. From our outreach efforts, we learned that many groups at DOTs will benefit from knowledge about existing AI resources, as they often struggle integrating these solutions into their programs. The last decade has seen the emergence of many AI-based solution providers. These companies enable easy integration of AI-based techniques related to software development, sensor technologies, data collection, data management, cloud computing, and data security. Some of these companies are dedicated to specific service areas like surveillance, supply chain management, asset management, and cybersecurity. Collaboration with these entities will help DOTs integrate the benefits of AI in the transportation application. Direct deployment of industry-based solutions will also minimize duplication of development efforts within DOTs.

Potential Benefits

• Identification of industry-based solutions will help minimize workload and management of internal workforce.
• DOTs will have access to diverse sets of solutions that will help them solve problems and make decisions quickly.
• Easy integration of AI-based solutions will benefit the public. This will accelerate the development of these solutions in multiple sectors.
• This effort will centralize a solution space that all entities across the country can access. Also, this may initiate effective collaborations among agencies.
• With a common platform, industry can also benefit by knowing the existing and imminent problems that require AI solutions.
• This collaboration will also initiate new research and development opportunities involving academia, industry, and government.

Objective²⁸

• Identify emerging industry stakeholders who provide AI-based solutions that can benefit DOTs for transportation research.
• Create a plan that could encourage partnerships between DOTs and the industry.
• Focus on building criteria that could aid DOTs in efficiently choosing an AI solution partner.

Tasks

• Identify example problem statements in transportation research. Example areas include, but are not limited to, traffic management, work zones, transportation systems management and operations, highway management and design, pavement, road safety, asset management, urban multimodal corridors, accessibility, travel behavior/behavior modeling, transportation equity, vulnerable road users, winter road operations, and commercial vehicles and freight operations.
  • Conduct a workshop with DOTs and partners.
  • Discuss current challenges in integrating the industry partners.
  • Develop a report summarizing current need for AI integration and expected requirements from the DOTs that an industry-based solution provider should deliver.

___________________

²⁸ https://www.whitehouse.gov/wp-content/uploads/2023/05/National-Artificial-Intelligence-Research-and-Development-Strategic-Plan-2023-Update.pdf.

• Identify commercial providers who offer transportation solutions to the problems identified in the previous task and integrate AI. The major AI-related areas may include computer vision, high performance sensors, big data analysis, advanced statistical analysis, ML, mathematical models and simulations, robotics, software design, human computer interface, and cybersecurity.
  • Conduct telephonic interviews with the industrial solution providers to identify their strengths and limitations.
  • Perform a literature review through websites, solution manuals, white papers, and technical reports.
• Develop a guideline for DOTs to select a solution provider.
  • Create a list of evaluation criteria.
  • Develop a model for cost benefit analysis.
  • Provide a list of selection criteria based on the DOT’s exact requirements.
• Develop a workforce requirement plan inside DOTs who can facilitate collaboration, identify partners, and evaluate solutions provided by industry partners involving AI.
• Develop a prototype of an interactive platform that will facilitate communication between these two groups.
• Conduct a workshop that will include industry, DOTs, and academia with a goal to:
  • Communicate imminent sets of problems that need focus.
  • Define requirements from DOTs for an effective solution.
  • Allow representatives from industry to define their capabilities and overlaps.
  • Form a pathway to effective collaboration.

Expected Deliverables

• A guideline for DOTs to select industry-based solutions for specific transportation problems.
• A workshop report on current scope and challenges in industry collaboration.
• A roadmap for effective collaboration.

Expected Budget

$400,000

Expected Duration

24 months.

Additional Resources

Partnership Development in the Federal Government: https://www.ida.org/research-andpublications/publications/all/p/pa/partnership-development-in-the-federal-government.

Research Problem Statement 10: Guidebook to Create Sharable, Reliable Sources of Data Sets

Background

The current revolution of AI is driven by large-scale data, and most modern AI models are data driven. However, there is a scarcity of reliable large-scale data sources in transportation research that can help solve problems at scale. Many DOTs already collect their own data using advanced sensors; however, these data are often targeted for very specific applications and geographic areas using different parameters, data dictionaries, and formats. The data collection efforts across DOTs are often disjointed. This can result in duplication of effort and hinder the transfer of knowledge. Due to budget limitations and scope, this also results in sporadic collection and insufficient data for modern AI applications.

The other major problem in existing data and collection efforts is the reliability and completeness of the data. Every AI method has its own requirement for data structure, minimum quality, scale, and associated metadata. Any AI operational system should follow standard practice for developing such data sets. It is important to create guidelines on collection, annotations, storage, and sharing of data.

The process to collect, manage, and share large-scale data needs proper supervision and coordination. The data require protocols for safety, security, accessibility, and governance. Such a process also needs to include protocol for standardization across sensor modalities, target applications, data annotations and metadata, and updates. Therefore, this project aims to develop a roadmap for such efforts so that DOTs can benefit from this standard protocol and accelerate AI adoption.

Potential Benefits

• Data standardization and sharing will accelerate AI adoption across DOT operations. For example, data sharing will help build prototypes more quickly.
• Enforcing reliable sources of data will also add benefit in refining future collection efforts.
• Transportation agencies often collect and manage data in separate silos, making it challenging to integrate and analyze information from different sources. This fragmentation can impede the development of comprehensive AI solutions. Successful deployment of this project can highlight key issues in this problem.
• Most modern AI systems depend on large-scale data. Often, data collected by a single DOT may not be sufficient to train large neural networks. Collaboration can solve that problem. Secondly, data collected in different states across the country will increase diversity of the data set, resulting in higher chance of generalizability, effectiveness in edge cases, and minimization of biases.
• Sharable data can help small DOTs that may not have the resources of larger DOTs to generate their own data, which can increase equity with access to the benefits of AI transportation data. Creating data in a standard format will facilitate sharing and collaborations.

Objective

• The goal of this research is to first identify already existing data sets along with the transportation research areas for which these data sets are applicable.
• The project will focus on selecting attributes that define data quality and provide a path for improvement in the existing data resources.
• The project will also identify the data gaps that exist in research. This will help to identify data needs across states.
• The project will identify necessary steps for data standardization, data governance, data sharing protocol, data privacy and security, metadata documentation, and data accessibility.

• Finally, the project will develop a roadmap for how to collect new data (including from industry partners), manage the data, and make data sets sharable across DOTs.

Tasks

• Perform a detailed review of existing data sets available for transportation research.
• Create a data needs report for different transportation applications (see report from NCHRP Project 23-12 to identify potential application areas).
• Create and execute a stakeholder engagement plan to discuss current requirements, current practice, existing protocols, and challenges in creating a sharable data set.
• Develop a technical document on data sharing protocol and accessibility along with possible challenges. Involve stakeholders as required.
• Finally, develop a roadmap for practices and protocols to facilitate creating sharable and reliable data sources for AI applications in transportation.

Useful Reference

• Results from NCHRP Project 23-12 Artificial Intelligence Opportunities for State and Local DOTs – A Research Roadmap include a list of AI tools, methods, and challenges. The literature review report also shows an initial list of dependencies between transportation problems and AI.
• International, D. (2017). DAMA-DMBOK: data management body of knowledge. Technics Publications, LLC.
• NCHRP Report 508: Data Management and Governance Practices.
• 20-44(48) Peer Exchanges on Data Management and Governance Practices.
• NCHRP Web Report 282 Framework for Managing Data from Emerging Transportation Technologies to Support Decision-Making.

Expected Deliverables

• A set of technical documents depicting the challenges and scope of creating practice and protocol for sharable and reliable data sources for AI.
• A data needs report for AI applications.
• A roadmap for sharable and reliable data sources for AI applications in transportation.
• A set of technical memoranda summarizing the data needs report and challenges.

Expected Budget

$350,000

Expected Duration

24 months

Research Problem Statement 11: Creating a framework to process and manage data collected by DOTs.

Background

The last decade has shown that data is immensely valuable for research and development in transportation. Due to advancements in sensor systems, electronics, and data storage facilities, we are witnessing an increasing trend in data collection efforts. However, any data collection effort requires structure to process and manage the data. DOTs and their associated agencies collect, manage, and analyze vast amounts of data to support their mission to ensure a fast, safe, efficient, accessible, and convenient transportation system. Some common types of data collected include traffic volume flow data, crash data, infrastructure testing data, roadway sensors data, traffic data, and many more. The sheer volume and variety of data pose several challenges for DOTs to manage and process the collected data.

• Efficient data storage is one of the major challenges that DOTs face. Since data is collected from various sources like road sensors, traffic cameras, and manual reports, the challenge is to ensure data quality, which requires continuous validation and cleaning.
  • Management of multimodal data is challenging. Data needs to be properly annotated with associated metadata.
  • The large volume of data also has the potential to pose privacy and security risks for DOTs.
  • Data integration can be complicated, especially when trying to correlate and analyze data across different transportation modes and jurisdictions.

Significant progress has been made in creating AI tools for data-driven advanced analytics. There are a few market-ready resources that can be integrated at various levels of data management within DOTs. Resources like data integration platforms could help DOTs and partner agencies integrate data from various sources in real-time. Literature also highlights the availability of tools that can be useful at various steps of data management. Data annotation tools are available for labeling data—such as text, videos, or time series data—for better understanding and use in decision making. Data visualization is another important part of AI that is relevant to track performance measures, understand the results, and for model evaluation. Tableau and Power BI are some of the market-ready solutions that provide interactive data visualization tools that can be standalone, web-based, and collaborative. Data storage and advanced processing services are also available through cloud-based solutions (e.g., AWS, Oracle). Even though technological advancements exist, DOTs do not have all the resources at their disposal to undertake various data management tasks in-house. It is even unclear in what ways data analysis and AI can be used, and which tools are most suitable to manage large volumes of data. Therefore, guidance is needed to understand which AI resources are readily available and which can be adapted by US DOTs.

Potential Benefits

The proposed project will create a guidebook for DOTs and other partner transport agencies that will consist of data management and processing plans. A guidebook can provide standardized protocols for data collection, storage, and processing, ensuring that data across different departments is consistent and comparable. Guidelines or framework can also help ensure that data security and can serve as a primary resource for training DOT personnel.

Objective

This study will help create a manual and identify resources and AI tools to help data engineers in understanding the type of information that is collected and how it can be analyzed. This project will also

create a guidebook that emphasizes human-AI interaction to ensure there are no ethical biases during decision-making.

Tasks

• Document current practice(s) the types of data required as well as what data is collected by DOTs and how data is stored by different agencies.
• Define steps and tools that can be used to label the data for data engineers at DOTs to better understand the information collected and how it can be used for analyses.
• Document the practices required to maintain machine-human interactions, as there are inherent biases in the datasets that DOTs might not be aware of.
• Prepare a final technical memorandum documenting the research process and results.

Useful References

• Results from NCHRP Project 23-12 Artificial Intelligence Opportunities for State and Local DOTs – A Research Roadmap include a list of AI tools, methods, and challenges. The literature review report also shows an initial list of dependencies between transportation problems and AI.
• International, D. (2017). DAMA-DMBOK: data management body of knowledge. Technics Publications, LLC.
• NCHRP Report 508: Data Management and Governance Practices.
• 20-44(48) Peer Exchanges on Data Management and Governance Practices.
• NCHRP Web Report 282 Framework for Managing Data from Emerging Transportation Technologies to Support Decision-Making.

Expected Deliverables

• A report on AI tools that can be utilized by the DOTs for data processing and management.
• A report on tasks that can be outsourced to third-party sources by DOTs due to lack of funds or personnel who can perform those tasks.
• A guideline that gives detailed information on each of the steps involved in big data management.
• A guidebook that can be used as training material for DOT personnel.

Expected Funding

$150,000

Expected Duration

18 months.