NYC Transit Authority Tests AI to Prevent Subway Delays

NEW YORK — The Metropolitan Transit Authority (MTA) has launched a pilot program with Google Public Sector to enhance subway system reliability using artificial intelligence. The initiative, dubbed TrackInspect, aims to detect rail defects before they lead to service disruptions.

Rob Sarno, assistant chief track officer at the MTA, has leveraged his 14 years of experience to help teach AI systems the sounds of failing infrastructure. This pilot project began in September 2024, focusing on retrofitting Google Pixel smartphones on select subway cars to gather critical data on track conditions. By capturing audio and sensor data, the program is designed to forecast potential issues and improve overall safety and efficiency.

“By being able to detect early defects in the rails, it saves not just money but also time – for both crew members and riders,” said Demetrius Crichlow, president of New York City Transit. The program represents a significant technological advancement in the MTA’s efforts to combat delays stemming from aging infrastructure.

AI has been gaining traction in public transit systems across major cities. In 2023, Aecom completed a similar project for New Jersey Transit, focusing on crowd management, while the Chicago Transit Authority implemented AI for security enhancements. In 2024, Beijing saw the introduction of mobile AI payments to alleviate ticketing lines.

TrackInspect, recently announced, started as a proof-of-concept initiative at no cost to the MTA, but the future expansion of the program and its associated costs remain uncertain. The MTA, the largest public transit system in the United States with 472 subway stations and 237 local bus routes, has previously partnered with Google on projects like Amtrak’s arrival and departure updates.

Despite its scale, the MTA continues to contend with service interruptions. In December 2024 alone, the transit authority reported over 42,000 delays, a stark contrast to Chicago’s average of just 200 monthly delays. The TrackInspect program is targeted at reducing such incidents.

From September 2024 to January 2025, MTA configured six Google Pixel devices on four subway cars. These smartphones captured a staggering 335 million sensor readings, 1 million GPS locations, and over 1,200 hours of audio. Deployed in a manner that ensures passenger privacy, the devices utilized advanced sensors to identify abnormalities that may indicate maintenance needs.

“We are trying to analyze vibrations and sounds that could signal issues, like a loose joint or a battered rail,” Sarno explained. His experience was vital in analyzing the AI’s findings, which were compared with inspections conducted by MTA track inspectors.

The integration of the TrackInspect system has demonstrated promising results. It was able to identify 92 percent of defect locations identified by MTA inspectors, showcasing the potential for AI to play a critical role in maintenance strategies. Furthermore, based on available data, some types of delays, particularly those related to braking and rail issues, declined on the A line during the trial period.

While the pilot program with Google may have concluded, the MTA emphasizes its commitment to pursuing technological solutions for ongoing challenges. Future partnerships with technology companies are expected as the agency seeks to enhance track inspection methodologies further.