Public transit is a complex system that involves a network of people, equipment, and facilities in a geographical space. Transit agencies have notably adopted artificial intelligence (AI) as a tool with great potential to enhance various aspects of operations, both strategically as well as in real-time, to improve efficiency, safety, and service quality. AI’s applications can be expanded toward transit signal priority systems, scheduling, route analysis, fleet and personnel optimization, vehicle automation, maintenance, transit safety, coordination with services provided by transit partners (e.g., transportation network companies and other micro-mobility modes), ticketing, and customer service (e.g., alternative trip planning when disruptions occur, or enhancing call center staff to respond to customer feedback and questions). This is shown to be the case in the existing use of different AI/machine learning technologies for transit, which include using reinforcement learning for coordinating operational control, or using supervised learning frameworks, like deep neural networks or language models, to monitor, analyze, and predict demand. In addition, generative AI, such as ChatGPT and Bard, could be used to enhance management efficiency in communication and coding.
For years, transit operators and agencies have collected and used data to analyze and improve service. This abundance of data includes second-by-second vehicle location data; sensors on assets that measure temperature, vibrations, and other variables; ticketing/fare gate information; and cameras monitoring operating status throughout the system, all of which could easily be leveraged for AI. This is particularly due to the nature of different machine learning models, which are unique in their ability to process diverse data types (e.g., numerical, image, natural language). Despite the noted potential benefits of AI, we have no structured knowledge of how it is adopted by transit agencies, who the adopters are, nor the extent of its benefits to agencies and transit users. There is also no catalog of opportunities to integrate AI into existing operations to enhance safety and service quality and to reduce costs (for example, could AI be used to address labor shortages in transit operations). Other questions include: What are the costs of adoption? What are the barriers to adoption? What are some of the associated risks of adopting AI, and how reliable are the outputs?
The research objectives are to:
- Study the deployment of AI in transit operations and provide insights into existing and potential use cases of AI, success stories, challenges, and barriers that deter AI deployment.
- Study the benefits and costs of AI adoption in transit operations.
- Develop and establish a framework of common definitions, guidelines, and standards for AI adoption in transit.
