NCHRP 23-12

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

Appendix F: Implementation of Research Findings and Dissemination Plan

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

Alejandra Medina², Abhijit Sarkar¹, Aditi Manke¹, Matt Camden¹, Tammy Trimble¹_, Gerardo Flintsch¹

¹Virginia Polytechnic Institute and State University
Blacksburg, Virginia
October 16, 2023
²FM Consultants, Blacksburg, Virginia

Introduction

This is a memorandum for National Cooperative Highway Research Program (NCHRP) Project 23-12: Artificial Intelligence Opportunities for State and Local DOTs – A Research Roadmap. As part of this project, the Virginia Tech Transportation Institute (VTTI) has developed 11 roadmap ideas and research statements, which have already been submitted to the Transportation Research Board (TRB). The goal of this Roadmap is to facilitate the integration of artificial intelligence (AI) into transportation research at Departments of Transportation (DOTs). In this document, we present the Implementation Plan and Workforce Development Plan as part of Task 5.

Problem Statements

Project Workshop

A two-day workshop was conducted via Zoom with participants from state and local DOTs, academia, and industry. The first workshop was held on two days, October 3 and 12, 2022. The objectives of the workshop were to introduce the project, describe the research project efforts completed to date, and discuss the development of a research roadmap that identifies and prioritizes research needs for incorporating AI in DOT work. The workshops included a discussion of previous research on AI in transportation over the last 10 years with a series of focused sessions to discuss topics related to current deployment and a future roadmap to include AI. Specific topics discussed included current and future focus areas of transportation development at DOTs, challenges in adopting AI-based solutions, sustainable workforce, and infrastructure development within DOTs for AI readiness, the readiness of AI, program evaluation, and third-party collaborations.

The second workshop was held on March 7, 2023, in order to discuss the development of the research roadmap to identify and prioritize research needs for incorporating AI in DOT work. Participants discussed current and future focus areas of transportation development at DOTs, challenges in adopting AI-based solutions, sustainable workforce, and infrastructure development within DOTs for AI readiness, evaluation, and third-party collaboration.

During the workshops participants provided input regarding project ideas to be included in the Roadmap for the different priority areas identified. Table 25 shows the project ideas that were initially proposed. They are classified according to the research areas presented in the workshops (workforce development, infrastructure development, readiness and evaluation of AI, challenges in adopting AI, current practices and prioritization, external collaboration, and equity, policy, and planning) as shown in the table below.

Table 25. Project ideas with corresponding research areas.

Based on further discussions and ratings input, the overall research statement was updated and the following research statements were selected to be included in the Research Roadmap. The research statements were selected to be included in the Roadmap based on the ranking priority. Table 26 shows the final research problem statements, objective(s), and areas.

Table 26. Final Research Problem Statements, Objectives, and Areas. (1) workforce development and infrastructure development, (2) readiness and evaluation of AI, (3) challenges in adopting AI, (4) current practices and prioritization, (5) external collaboration, and (6) equity, policy & planning.

Implementation Plan

The implementation plan described herein encompasses the following: (1) activities already executed as part of the project, (2) activities slated for completion prior to the project’s conclusion, and (3) activities set to be finalized after the project concludes. The project team has identified five action areas for the dissemination plan:

• Awareness of the research conducted as part of the NCHRP 23-12 project
• Identification of potential stakeholders
• Development of educational materials
• Creation of a repository
• Implementation success performance practices

This document outlines the strategies aimed at enhancing awareness of the project. It identifies potential stakeholders, including national and state organizations. These stakeholders are envisioned to play a crucial role not only in shaping the Research Roadmap but also in disseminating the information garnered throughout the project. The memorandum also underscores the development of supplementary educational materials to bolster awareness. Additionally, it proposes the establishment of a repository and methodologies for identifying and quantifying the impacts associated with these implementation action areas. These areas of action exhibit variances in terms of implementation levels, timelines, and the potential requirements for additional funding.

The actual implementation schedule of the proposed project elements will depend on specific priorities of states and federal agencies, available funding levels through state and federal sources, and activities from partnering agencies like NCHRP, AASHTO, the Federal Highway Administration (FHWA), and others. Priorities may also be impacted by other aspects such as advances in technology. It is important to note that the problem statements presented can be modified based on the needs and goals of the supporting agency.

The proposed Research Roadmap and Implementation Plan comprehensively address a wide spectrum of AI applications for transportation needs. The foundation of this plan rests on the assumption that collaborative efforts among entities like NCHRP, AASHTO, FHWA, state and local DOTs, academic institutions, and the industry will underpin research activities, guide development, facilitate outreach, disseminate findings, and operationalize novel guidance and methodologies. Cooperation is essential for the successful execution of the initiatives outlined in this plan and is important not only to its success but also to the sustained integration and evolution of AI across diverse transportation undertakings.

Awareness of Conducted Research

The audience for this research will be broad and include state and local policy makers, academia, and private sector consultants and researchers. Since the project objective is to accelerate AI technology in state DOTs, it is expected that implementation strategies will focus on state and local DOTs and will be accomplished by working with AASHTO. The implementation will build upon completed and ongoing activities. As part of Workshop 2, we already presented these ideas to personnel from state, federal, and local DOTs.

Dissemination of Findings Through Conferences

One additional and effective way to disseminate the findings of this work is to reach out to stakeholders, researchers, and practitioners through conference presentations and journal articles. Dr. Sarkar presented the interim findings as part of discussion panel at the TRB 2023 Annual Meeting, “Are We There Yet? Discussing Applications of Artificial Intelligence and Machine Learning,” organized by the Standing Committee on Information and Knowledge Management (AJE45) and Standing Committee on Artificial Intelligence and Advanced Computing Applications (AED50). The title of the presentation was “Pathways

of Artificial Intelligence and Machine Learning in Transportation Organizations.” The team also submitted the findings from Task 2 for presentation at the TRB 2024 Annual Meeting.

The outcomes of this project are also expected to be shared through papers presented at future TRB annual meetings and various national and international conferences, including conferences hosted by the primary stakeholders.

Outreach to Stakeholders

This activity includes coordination with and reaching out to TRB Committees, AASHTO Committees, and other Transportation Stakeholders

Coordination with TRB Committees

Throughout the project, the research team has maintained correspondence with several TRB committees including:

• AED50 Standing Committee in Artificial Intelligence and Advance Computing Operations
• ACP15 Intelligent Transportation Systems
• AP020 Emerging and Innovative Public Transport and Technologies
• AED30 Statewide/National Transportation Data and Information Systems

In addition to these committees, several other committees in specific research areas were contacted (i.e., safety, infrastructure). During the 2023 Transportation Research Board’s Annual Meeting, the research team prepared a short presentation and a handout that was distributed in several committee meetings and panels.

Additional AASHTO committees to reach out as part of the implementation plan include among others:

• • AJE45 Committee on Information and Knowledge Management,
  • AJE70 Committee on Data for Decision Making,
• AJE15 Committee on Workforce Development and Organizational Excellence

Coordination with AASHTO Committees

AASHTO Committees have the capacity to play an important role, not only in the dissemination of the research but also in performing some of the activities listed in the research problem statements. As part of the implementation AASHTO committees must receive the products of these research. The following committees among others must be contacted:

• Data Management & Analytics Committee
• Knowledge Management Committee,
• Human Resources Committee,
• Agency Administration Committee,
• Policy, Program Delivery Committee
• Cross-Discipline committees.

Coordination with other transportation Stakeholders

There are numerous professional organizations like TRB, such as the Institute of Transportation Engineers, the National Association of City Transportation Officials, the National Association of Counties (NACo), and the National Association of Cities, among others, that can provide valuable support for the implementation of this Research Roadmap.

For instance, NACo recently launched an Artificial Intelligence Exploratory Committee in May 2023. This committee is dedicated to examining emerging policies, practices, and potential applications of AI. It represents county elected and appointed officials from across America and will focus “on the lens of county governance policies and practices, operations and constituent services, public trust, privacy and security, and workforce productivity and skills development.” Similarly, TRB AED50 is actively working in bringing AI into all disciplines of transportation.

Distribution of Final Deliverables

After final review by the panel members of the project, we recommend that the final report and accompanying Research Roadmap be distributed to those individuals who participated in the project interviews and workshops.

Identification of Stakeholders

As mentioned, the foundation of this plan is rooted in the efforts of NCHRP and AASHTO. Additionally, it relies on an assumption of collaboration with other organizations, including FHWA, state and local DOTs, academic institutions, and industry stakeholders. These collaborations will serve as the bedrock for research activities, guide development, facilitate outreach, disseminate findings, and operationalize innovative guidance and methodologies. The team has identified a group of stakeholders who may play a vital role in the implementation of the Roadmap.

The below-mentioned organizations and stakeholders each have distinct agendas concerning the advancement of AI in transportation. It is anticipated that the research endeavors undertaken in these projects will propel the progress of each organization’s objectives. Furthermore, it is inferred that the collaborative spirit between these agencies will transcend the boundaries delineated by the scope of this plan.

In addition to NCHRP and AASHTO the primary organizations expected to participate in the plan’s implementation or offer support include the following:

• State and local DOTs are the primary intended audience for this research. They are also the ones responsible for implementing noteworthy practices and policies. Individual efforts by the state and local DOTs, and collaborations among them and other institutions, are paramount for realizing the Research Roadmap developed in this project.
• AASHTO Committees holds significant potential as a vital partner for research implementation and funding across various activity categories. This potential arises from its sponsorship of the NCHRP, endorsement of national policies and manuals, and its participation in committees.
• United States Department of Transportation (USDOT) focuses on active transportation research that aligns with the respective missions of its administrations. The USDOT is “committed to safety and innovation and sees artificial intelligence (AI) as a promising capability to help achieve these aims.” (https://www.transportation.gov/) AI. The USDOT, and specifically FHWA, has put a special focus in the application of AI to improve not only research but also facilitate different transportation systems, including NLP, computer vision and ML-based prediction. The FHWA assumes a pivotal role in disseminating, executing, and integrating research discoveries and methodologies among states, metropolitan planning organizations, and local agencies. Sustained engagement from these offices stands as a crucial factor in propelling progress and effecting tangible improvements in introducing AI in different transportation processes. While most of the problem statements included in the Research Roadmap are related to FHWA activities, other branches of the USDOT, like the Federal Motor Carrier Safety Administration (FMCSA) and the National Highway Traffic Safety Administration (NHTSA), can benefit from this research and the proposed plan.

• Local Technical Assistance Program (LTAP) / Tribal Technical Assistance Program (TTAP)
• LTAP and TTAP Centers provides not only outreach but also training to local governments and Tribal Agencies in all 51 States and as such can be key to reach an audience not covered but other organizations.
• Academia will play a special role in the adoption of AI in different transportation projects. AI is an area that needs special skills that for the most part is not taught in traditional engineering railroad innovation programs. Academia can fill a serious gap in technical capabilities and building fundamental understanding. Based on our interviews and workshops, states and local DOTs are already building strong relationships with academic institutions. These connections help them study the credibility and usefulness of novel AI systems and include them in their solution pipeline.
• TRB Committees already play, and will continue to play, a crucial role in identifying emerging AI applications and facilitating outreach initiatives. These outreach activities include webinars, workshops, and conferences that spotlight recent research tailored to practitioners’ needs.
• Metropolitan Planning Organizations (MPO). MPOs are the policy board representing localities in all urbanized areas with population over 50.000 and as such several of the research problem statements produce as part of this research will be of special interest to them (Workforce Development)
• Professional and nonprofit organizations can be an excellent conduit to communicate research results and promote collaboration.
• Private organizations and industry are vital to develop AI programs and products that meet the needs of the transportation industry. Transportation processes rarely are unique to one state. That makes it an attractive market for private organizations and industry to create products that include AI that can be adapted for the needs of each state.

Table 27 details potential partnerships, projected budget, and estimated duration for each research problem statement outlined in this Roadmap. These potential partnerships can involve funding, leadership, support (such as data collection and provision), or assistance in implementation and outreach. Some projects may also receive strong backing from one or more states. While this may result in a more confined scope, it is anticipated that the outcomes will hold substantial value for nationwide dissemination.

Throughout the interviews and workshops, it was noted that several ongoing projects align with FHWA research activities. Consequently, in addition to USDOT, FHWA has been identified as a specific potential partner for this research initiative.

Table 27. Problem statement potential partnerships and expected budget and duration of each research problem statement proposed.

Development of Education Materials

The creation of educational materials will complement the previous work to raise awareness about the research. The proposed materials possess the flexibility to be utilized in diverse settings and cater to a range of stakeholder interests. This audience includes DOT administrators, subject matter experts, and transportation professionals in general. These educational resources collectively contribute to a multifaceted approach for disseminating knowledge and fostering understanding of the subject matter.

The research team suggest the following as potential educational material for distribution:

1. 10-minute buy-in presentation material: This concise presentation will be aimed at decision-makers. Its purpose will be to provide an overview of how AI can enhance various transportation processes that fall under the purview of DOT responsibilities. Additionally, the presentation will offer insights into the AI-related activities of different states and introduce the suggested problem statement.
2. 45-minute stakeholder presentation material: Tailored for different stakeholders, this comprehensive presentation will cover the current state of AI in transportation. It also will incorporate findings from interviews and workshops, along with the roadmap. The objective of this presentation will be to equip practitioners with a deeper understanding of the challenges and

2. limitations associated with transportation AI strategies. This presentation will serve as the foundation for the subsequent webinar.
3. Webinar: Tailored for practitioners, the objective of this webinar is not only to create awareness of the research, but also to encourage collaboration among the different entities. It is expected that professionals from one or two states will be invited to share some of their experiences with AI implementation.

Creation of a Repository

While the final roadmap problem statements marked a significant milestone in NCHRP Project 23-12, the research team considers it important to establish a process for maintaining and updating the Roadmap, transforming it into a dynamic and valuable living document. The continuous evolution of AI in transportation will necessitate considering new developments and experiences. Special attention needs to be placed to the Roadmap’s inherent crosscutting nature (encompassing multimodal, multi-user, and multidisciplinary aspects).

There can be various potential avenues or initiatives for the sustained advancement of the Research Roadmap. Options used in the past is to designate a TRB or AASHTO committee or include in the research activities of TRB, FHWA or others a project to update the present roadmap at regular close intervals. However, based on previous experiences addressing roadmaps in areas of extremely broad applications—where AI is considered more a tool than a specific area and involves several disciplines, and considering the rapid changes in AI technologies compared to other areas, along with the fact that several TRB committees are already proposing AI projects—it seems that the best approach is the creation of a task force. This Task Force can be led as a joint or separate effort by TRB, AASHTO, FHWA, or the states. It is imperative that this Task force includes all the stakeholders mentioned in this document, especially the industry and academia.

The objective of this group or task force will be to identify critical issues associated with artificial intelligence that state and local DOTs will face and provide recommendations for potential funding. To facilitate the process, two or more subgroups can be created in specific areas. The recommendations of this Task force, expected to be issued at least on an annual basis, will be taken by committees and other stakeholders to propose or conduct research to address those issues. In order to better served the states a committee or sub-task force must be created under AASHTO or TRB’s umbrella to study how the Task Force recommendations have been addressed, identify areas that were not covered, and identify mechanisms to do so.

A more complex and appealing alternative is to create an NCHRP Task Committee or Task Order support project where the objectives will not only be to identify critical issues but also to conduct research on those issues (including an update of the roadmap) and conduct related technology transfer. This approach will require a designated fund source and must cover several areas or create a task force under AASHTO or other organizations. As of the time of this writing, no funding has been allocated for this task.

Challenges Affecting Potential Implementation

Barriers and challenges for accelerating the adoption of AI by state and local governments were identified by the interviews, workshops, and literature review. It is important to note that most of these challenges are

similar to the challenges encountered to adopt AI generally and are not specific to transportation. (https://rosap.ntl.bts.gov/view/dot/66971/dot_66971_DS1.pdf).

Major challenges in the implementation of AI include:

• Trust in the capabilities of AI
• Integration with existing systems
• Lack of sufficient high-quality and relevant data
• Supporting funding and technology needs
• Potential Bias
• Lack of workforce expertise
• Model Extrapolation

To mitigate the impacts of potential challenges, the research team created a register covering the main risks. The register includes management actions that could be used to mitigate each risk. The risks identified within this section are classified with ratings for three aspects of each individual risk: the probability of that risk occurring; the impact on the project cost, schedule, or scope; and the ability of that risk to be mitigated. These levels are defined in Table 28 (Note: risks were rated using the Intelligent Transportation Systems Joint Program Office standard, which may be found at https://www.its.dot.gov/project_mang/index.htm). Table 29 summarizes the main challenges expected when implementing the benefit-cost analyses framework and the experienced-based strategies for mitigating a risk’s potential impact on the project. Risks are identified using a taxonomy that includes institutional, personnel, and technical risks. Table 29 also lists the challenges their ratings and probabilities, and planned mitigation strategies.

Table 28. Risk rating and probability definitions.

Table 29. Challenge matrix.

Workforce Development

The lack of a skilled AI workforce was mentioned by the states as one of the major hurdles to overcome in order to accelerate the integration of AI methodologies into transportation-related applications. This skilled workforce can originate within the state and local DOTs, industry, or the academy, and, in general, hinges on a comprehensive training strategy aimed at both the broader workforce and specifically targeted at state and local DOT personnel.

Based on state interviews, workshops, literature, and state of the art analysis, the research team has identified workforce challenges that states have already encountered or anticipate encountering when implementing AI strategies. Furthermore, several agencies mentioned the expectations from upper management that the agency as a whole supports the application of AI technologies to create better solutions to transportation problems. However, in doing so, management tends to underestimate the need for resources in general and the workforce in particular. This lack of workforce expertise within DOTs related to AI and ML restricts states’ abilities to promote AI projects.

As with every initiative involving state, private, and non-profit associations, there are pros and cons to contracting the private industry or academia to fulfill the needs related to state AI projects. However, there is agreement among DOTs that the success of AI implementation in transportation will depend, among other factors, on how to find and work with other partners, especially industry and academia. Potential benefits of contracting AI projects include among others:

1. Having the ability to incorporate state-of-the-art experience in AI to develop the products needed by the states.
2. A relatively direct and quick process to get the project done efficiently and on time.
3. No need to hire new experienced employees or train new ones.
4. Potential reduction in costs if the same approach can be applied to different state and local governments.
5. The ability to transfer the AI responsibility to the consulting firms.

Additionally, potential cons of contracting AI projects include, among others:

1. Limited knowledge by AI consulting firms of state transportation strategies, as generally, most consulting firm employees are experts in AI but not in transportation. This may result in projects that do not maximize benefits or fail to reach the goals of the projects.
2. Limitation of the AI products delivered for the specific area, making it unable to use some of the processes developed in other areas, resulting at the end increasing costs when several projects are considered.
3. Difficulty in reaching a level of understanding with the consulting firms about the real needs and goals.
4. Since AI is expected to be incorporated into future projects, a lack of expertise can result in the inability to effectively manage the contract.

The success of the AI transportation project will require a delicate balance of all the stakeholders involved. However, there is also agreement among states on the need for a strong state DOT counterpart

workforce throughout the process. State DOTs have previously faced crosscutting workforce challenges. However, there is need for structured strategies. These strategies may include

• Access to Educational material: To maximize the benefits associated with the identification and implementation of transportation-related AI projects, there is a need to identify the required workforce skills, education, and training. In the near term, this could include identifying the necessary skills, developing skills courses, and making the courses available to DOT personnel. Given the diversity of the research at DOTs, these courses need to be diverse in nature. Starting from fundamentals to specific application-focused courses will help the DOTs.
• Outreach: 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. For the optimization of outcomes arising from the conceptualization and execution of AI-driven projects in the transportation realm, it is paramount to define the essential proficiencies, educational requisites, and training pathways. In the short term, this could involve identifying pertinent competencies, formulating educational modules, and facilitating access to these resources for DOT personnel. Furthermore, active participation of DOT employees in national and local AI-focused platforms should be actively encouraged, fostering the exchange of insights and know-how both within and between different state DOTs.
• Identifying key roles: The General Services Administration (GSA) has developed a living and evolving guide for the implementation of AI applications for the US government that can be adapted as a first step for state and local governments (Table 30). Please note that the positions identified by GSA refer specifically to AI expertise, but AI applications also require personnel in other specific domain expertise’s like data science, cybersecurity, and computer science. Furthermore, it is extremely important for the organizations to have professional experts in the different AI application areas that can clearly communicate the needs and expected outcomes to the AI experts.
• Workforce retention and recruitment: With the changing demography and landscape of applications, new workforce needs to be introduced to the DOT’s work plan. Universities are increasingly offering interdisciplinary research where skills from computer science, electrical engineering, and AI are integrated with civil engineering, mechanical engineering, material science, management studies, and other relevant areas that conventional transportation engineers are experts in. DOTs should create resources and scope for the new graduate who will bring this knowledge and expertise. At the same time, the DOTs should identify members from the existing workforce who are able to imbibe the new knowledge, upgrade their expertise, and channelize knowledge into efficient deployment.

Table 30. Typical roles for AI personnel (Source: GSA, 2022).

Developing Workforce

The following steps are recommended to define a strong, reliable, and equitable AI workforce:

1. Identify in the organization people that already have the necessary skills. The successful implementation of AI strategies in existing and new processes requires people who use data regularly to explain decisions, are comfortable with data analysis process, have demonstrated that they are eager to learn new technologies, and in the past have promoted the adoption of new technologies to improve existing processes or create new ones. As a result, the identification of the potential candidates must not be limited to people who have AI skills, and it is recommended to extend the pool to include people or teams with strong analytical skills
2. AI implementation plans based on the mission, short-term goals, and long-terms goals:
   • Identify the type and number of positions needed for the AI implementation process.
   • Create the job description for each of those positions.
   • Identify the skills needed for each position with special attention on how candidates will be evaluated.
   • Determine the costs of including these employees on the organization roster.
3. Identification of potential partnerships with academia or industry. Academia and industry have been for years strong resources when there was a need to implement AI processes. Identification of potential counterparts in academia is useful when (1) it is necessary to complement an existing skilled workforce in the organization, (2) the application requires highly skilled workers in the short term but not in the long term, or (3) the organization wants to test an application but is not sure if the AI application will be adopted in the long term. Associating the DOT personnel with such projects will help build the technical know-how, comfort, and trust in the AI system.
4. While 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

4. counterpart workforce throughout the process. State DOTs have previously faced crosscutting workforce challenges. Strategies to address these challenges include:
   • Working with partner agencies and other stakeholders (e.g., contractors, industry partners, and the general population);
   • Establishing working groups focused on workforce development activities;
   • Supporting employees as they pursue the acquisition of new skills.
5. Developing and implementing training programs. AI training is critical for the success of AI implementation. Actions under this program development include:
   • Work with industry stakeholders to identify and catalog available courses and publish them internally.
6. Development of staff recruitment and retention programs that respond to current and future needs of AI implementation strategies. To promote retention and make any AI roles more appealing, it is important to identify positions and roles that really contribute to the mission of the organization to provide safe, efficient, sustainable, and equitable transportation for all users. A first step is to create a competent leadership team across different levels at DOTs. The leadership team should define the overall goals and strategy so that top-down communication is clear, the role of each individual is defined, and the overarching destination is well understood.

Conclusion

This deliverable summarizes a dissemination plan for the Research Roadmap developed as part of project NCHRP 23-12. The Research Roadmap, which was delivered as a separate deliverable, proposes a roadmap that can be used to efficient implementation of AI-based methods inside DOT. This document lists a set of next steps that the performing team is developing to disseminate this project’s outcomes. The team also lists suggestions and recommendations that the sponsor and other stakeholders should follow for successful dissemination of the Roadmap. We have also highlighted key challenges, risks, and steps to mitigate the risks while deploying the Research Roadmap in general at DOTs. The Research Roadmap was created in consideration of diverse needs and diversity in application scopes and workforces at DOTs. Therefore, following the proposed steps will also help all levels at DOTs to strategize.