APPENDIX B
Annotated Bibliography

Mulholland, I. P., and R. A. Oren. 1994. TCRP Report 1: Artificial Intelligence for Transit Railcar Diagnostics. TRB, National Research Council, Washington, DC. https://www.trb.org/Publications/Blurbs/153845.

This report was written for transit railcar maintenance professionals who were concerned with improving railcar maintenance fault-diagnostic capabilities by using artificial intelligence (AI) technologies. This report serves as an early example of TCRP investigating AI technology for the use of diagnostics; it also serves as a historical reference for this topic. The report defined AI as “a computer program that uses human problem-solving techniques to assist and augment the diagnostic process.” The report investigated seven AI technologies with the goal of determining their potential as related to their application for transit railcar systems and subsystems.

The AI technologies that were investigated included:

1. Expert systems,
2. Case-based reasoning,
3. Model-based reasoning,
4. Artificial neural networks,
5. Computer vision,
6. Fuzzy logic, and
7. Knowledge-based systems.

The report concluded that AI technology was sufficiently mature for application in the transit railcar diagnostic process and that it would be cost-effective. It also provides recommendations for implementation of the AI technology that was investigated.

Caltabiano, R., R. Camus, R. Gerin, and G. Longo. 1997. Implementation of an Advanced AVM System for the Trieste Bus Network. University of Trieste, Department of Civil Engineering.

This report explained why AVM systems are useful for improving the quality standards of the public transport service. It stated that the improvement may have been related to the increased appeal of public transportation at the time this article was being written. Research concluded that through use of AVM, it was possible to better realize public transportation improvements (as they related to modal split and reduction of costs) while increasing commercial speed. Like many AVM evaluations of that period, it focused on service and operations improvements (including on-time performance), rather than on AVMʼs capacity to provide quantities of vehicle health and diagnostic information.

Schiavone, J. 2010. TCRP Synthesis 81: Preventive Maintenance Intervals for Transit Buses. Transportation Research Board of the National Academies, Washington, DC. https://doi.org/10.17226/22965.

This study (TCRP Project J-7, Topic SE-05: “Synthesis of Information Related to Transit Problems”) was written to educate and assist transit administrators, engineers, and researchers in the face of potential issues for which useful information already exists. It noted that there is information on a multitude of concerns within the transit industry, that expensive research findings could go unused, and that valuable experience could be overlooked.

Most preventive maintenance findings came from the work of practitioners faced with problems in their day-to-day work. The TCRP Oversight and Project Selection committee authorized a study to provide a system for the assembly and evaluation of such useful information and to ensure that it is made better available to the transit community.

The purpose of this study was to find and synthesize effective information from available sources and to prepare concise and documented reports on preventive maintenance of buses. Although it focuses on conventional methods for establishing preventive maintenance scheduling, the report points to the potential for vehicle monitoring and diagnostic systems in the following lessons learned:

Nessen, S. 2022. MTA to Use Artificial Intelligence Tech to Keep Buses from Breaking Down. Gothamist. https://gothamist.com/news/mta-to-use-artificial-intelligence-tech-to-keep-buses-from-breaking-down. Accessed April 10, 2023.

In this press release, the Metropolitan Transportation Authority, a public transportation organization in New York City, describes its plans to use AI technology to help prevent buses from breaking down. For two years prior, the agency tested technology from Preteckt. The technology was described by the companyʼs owner, Ken Sills, as being like a “check engine” light turning on in a car when something is wrong. He continued to describe the way that the technology can identify potential serious equipment problems before they take place, which allows for maintenance to be performed proactively. This means that smaller, less expensive issues can be fixed before they turn into more damaging and costly ones.

Ray, R. A. AI and Transportation. 2023. Eno Center for Transportation. https://enotrans.org/eno-resources/ai-and-transportation/. Accessed June 14, 2024.

This article discusses how AI can impact the transportation sector. The articleʼs definition of AI is the same as that used by this study when key terminology was defined at the beginning of the report (see Chapter 1). The author describes AI as rapidly developing in its ability to automate repetitive tasks and expand the reach of data collection and analysis. AI is increasingly likely to be incorporated into the work of public transportation agencies, especially as they continue to build and maintain digital assets as well as physical ones. As a result, it is necessary for practitioners to understand how AI works—including the benefits and risks that could be associated with utilizing this technology.

This article lays out a variety of available resources that explain how AI can be used in transportation—specifically, the practical considerations that transportation professionals and policymakers should be aware of. The article also addresses gaps that can be found when using AI. The author concludes that if “thoughtful, well-informed transportation experts help guide AI development, they can dramatically improve the quality of AI tools and maximize the benefits to the industry, helping create a safer, more efficient, and more equitable system.”