Artificial Intelligence in Public Transit: Better, Faster, Safer?

Expanded laws, grant funding, and new technologies now make it possible for California agencies to invest in AI systems. What’s available, and what’s at stake? 

By Arianna Smith
Managing Editor 
Transit California 

Thanks to recent law changes and technological advances, California transit agencies have new opportunities to invest in artificial intelligence (AI) as a tool to reduce traffic congestion, travel times, and costs, improve customer experience and system safety, and increase revenues.  These new legal authorizations and products have the potential to increase agencies’ ridership levels, but individual agencies are carefully analyzing costs, safety, and privacy concerns when choosing whether to and how to move forward with AI.

With the explosion of AI products and services offered across nearly every policy sector, a common understanding of what AI is and does specifically within the transportation sector can help inform transit decisions. The International Association of Public Transport (UITP) 2020 report, “Moving Forward with Artificial Intelligence in Public Transport,” explains that AI is not universally defined, but for AI applications in public transportation, the technology should have the ability to learn from and adapt to data, to show creativity in solving issues, and to improve processes that already exist. 

According to the report’s surveyed organizations, the most frequent AI uses were for real-time operations management and customer analytics (25% of organizations surveyed each used AI for these functions), intelligent ticketing systems (21%), predictive maintenance (17%), scheduling and timetabling (17%), multimodal journey planning (17%), disruption management (15%) and fraud detection (13%).  Ten percent or less of surveyed organizations used AI for safety management, network planning and route design, mass customization service, customer support, and/or administrative tasks.

The US federal government has supported a variety of AI projects at transit agencies around the country through grant funding. In 2020, as part of a $130 million grant for advanced vehicle technologies research, the US Department of Energy’s Vehicle Technologies Office and the US Department of Transportation’s Federal Transit Administration jointly funded Tennessee’s Chattanooga Regional Transportation Authority (CARTA) to improve energy efficiency and affordability in its public transit system. Using the grant, CARTA partnered with Vanderbilt University and the AI company SmartTransit AI to develop an AI software platform that analyzes rider demand, traffic congestion, and vehicle energy use. The pilot project begins in March 2024, and will use “artificial intelligence and real-time data analysis at scale” in order to develop “models to estimate the load factors and real-time energy consumption of mixed-vehicle transit fleets and use those models to predict and optimize operations in order to lower overall energy impact while ensuring that system-wide capacity remains unaffected,” according to the SmartTransit AI website. A key goal of the program is for AI to design the most energy-efficient blend of fixed-route transportation and on-demand transportation for riders.

Some researchers in California are focusing on applying AI to the state’s established general transportation priorities. Under the Mineta Transportation Institute (MTI), California State University Transportation Consortium (CSUTC) researchers are currently developing a project to save the lives of cyclists and pedestrians, which made up 43% of traffic fatalities in 2021 and, while levels are now slowly falling, continue to be a major cause of preventable death in the country. In spite of this, California lacks an automated system that can detect or predict these risks in urban areas. CSUTC’s AI-powered system could fill that critical gap “to monitor, measure, and control the traffic flow particularly at intersections, and detect and identify Near-Miss Collisions for pedestrians and bicyclists as an important indicator to recognize and measure actual risks.” The research project, which advances the National Association of City Transportation Officials' Vision Zero initiative for cities to develop safer streets, runs through May 2024.

Another major discussion for current AI use in California transit systems focuses on the enforcement of a commonly used transit innovation that has proven critical to easing traffic congestion in urban areas: transit only traffic lanes. The April 2023 edition of Transit California’s feature story, “An Update on Ridership Recovery,” describes many of the Association's member agencies’ installation of these specially marked preferential-use lanes to improve speed and reliability of their systems as part of the larger efforts to attract and retain transit riders. However, transit-only lanes aren’t separated from regular traffic, so non-transit vehicle drivers sometimes illegally drive or even park in these lanes. For decades, transit agencies around the state have sought cost-effective enforcement to deter illegal lane use.

A 2007 pilot program allowed the San Francisco Municipal Transportation Agency (SFMTA) to use forward-facing cameras for parking control on transit buses in order to enforce parking violations in transit-only traffic lanes and transit stops, and a 2016 law that expanded the program to AC Transit. The programs worked: According to a 2021 California State Assembly Floor Analysis of later legislation, “A report SFMTA issued on the effectiveness of that program showed that during the pilot phase of the program, transit delays were reduced by as much as 20%.” AC transit, which began its program in 2020, “reported that on-time performance improved from 54% to 75% since the service began.” Enforcement activity is also intended to improve the safety of and accessibility for bus boarders, who rely on transit stops to be clear and available only for authorized transit vehicles.

The 2021 law, Assembly Bill 917 authored by then-Assemblymember Richard Bloom (D-Los Angeles) and co-sponsored by the Association, LA Metro, and AC Transit, expanded the authorization to all of California’s public transit operators who passed resolutions to install automated cameras. The law requires agencies who install cameras to provide a report to the Legislature by 2025 about their system’s effectiveness, costs, and revenue generation, as well as on the impact to privacy and traffic control. LA Metro is the latest and largest agency to install these cameras on some of their buses for transit only lanes and bus stop enforcement, which will be in service later this year.

The automated cameras in LA Metro’s buses will use AI technology made by Hayden AI to identify vehicles violating lane and transit stop laws. The cameras will take pictures of law violators’ license plates, and law enforcement will send them warning notifications or tickets.

“The mission of Hayden AI is to use computer vision and geospatial technologies to make cities safer, more sustainable, and smarter. Partnering with LA Metro to keep bus lanes and bus stops clear helps advance that mission,” said Chris Carson, CEO and founder of Hayden AI, a Business Member of the Association. “[Our] technology makes public transit faster, more reliable, safer and efficient. We’re excited to help deliver high-quality transit service for Angelenos.”

Still, some organizations remain concerned about AB 917 and other measures that authorize AI applications in public settings. The ACLU and Electronic Frontier Foundation, which opposed AB 917, wrote, "While this measure has a worthy aim to decrease obstructions in transit-only lanes and make public transportation more efficient, it seeks to do so by relying on the false notion that increased surveillance is the solution. Californians are increasingly concerned that Big Tech, government agencies, and the police have collaborated to create a vast surveillance infrastructure that extracts our personal information for their own profit and power. Despite your good intentions to help prevent obstructions to efficient public transportation, AB 917 goes too far by fostering the use of surveillance technology that serves its founders and funders at the expense of the rest of us."

The AI-powered camera systems and the California agencies that choose to deploy them are subject to numerous codified rules intended to address privacy concerns raised during the measure’s discussion. Agencies that plan to install cameras must issue notices months in advance, and provide warnings rather than tickets for certain time periods. Cameras are required to be installed at angles that capture only the license plates of vehicles violating parking rules rather than other more general activities occurring on the street. Also, videos must be destroyed within a certain time frame.

Broader concerns over AI use in transit exist as well. The 2020 UITP report addressed the possibility of overreliance on AI: “Like any technical systems, AI‑based systems are prone to hacking or denial‑of‑service attacks, possibly by other AI systems developed with malicious intent. The risk is when public transport operators have [rationalized] the manpower with overreliance on AI, which led to abandonment in training to manage situations previously handled by human intervention.”

At an expert panel at the LTA-UITP Singapore International Transport Congress and Exhibition in 2018, panelists brought up further AI limitations that must be considered when using AI in transit: “AI bias is a well-known problem in some consumer applications. For public transport, a possibly greater issue is that the segments of the population who are not producing data using their mobile phones (elderly people, primarily) may be hard to reach using dissemination mechanisms to be able to leverage the benefits of the AI.”

Unquestionably, AI opportunities will continue to expand across every component of public transit systems. The benefits to some AI-powered or informed systems are well studied and have clearly established their ability to increase efficiencies, improve safety, and reduce costs. Other applications must continue to be researched, further developed, tested locally, and carefully analyzed before wider adoption. The Association is prioritizing AI policies that best uplift and secure the future of California’s public transit systems.

Get the insider’s perspective

Stay up to date on everything important in transit today.

Subscribe now
Connect with us