AI is telling New York subway workers if that suspicious sound is a problem
The Metropolitan Transportation Authority, responsible for managing public transit across New York City, launched a pilot program last Thursday that uses a prototype developed by Google Public Sector to detect potential track issues by using sensors and artificial intelligence.
Called TrackInspect, the program attached Google Pixel consumer smartphones with built-in sensors and attached microphones onto subway cars in order to capture vibrations and sound patterns — such as suspicious rattles, bangs or squeals — as trains moved throughout the subway system.
The data collected from the devices, in addition to MTA’s existing track information, allows subway track inspectors to address maintenance issues, including equipment failures, trash buildup, or construction, before they disrupt operations that can cause train delays and public safety hazards.
“The TrackInspect pilot is a game-changer for the MTA, combining advanced cloud, AI, and real-time sensor technology to transform how we maintain and monitor our subway infrastructure,” Rafail Portnoy, MTA’s chief technology officer, said in a press release. “It reflects our commitment to uniting technology and operations to drive innovation and safety.”
The initial pilot was conducted on six cars on New York’s A train subway line, which runs from upper Manhattan to Queens, from September 2024 to January 2025. TrackInspect collected 335 million sensor readings, one million GPS locations and 1,200 hours of audio, according to the results presented by the Google team at an event last Thursday. The sound and vibration data was combined with the city transit authority’s database and shared with a machine learning model running on Google Cloud. It was subsequently reviewed by MTA track inspectors to confirm whether the system correctly identified issues.
TrackInspect also uses generative AI so subway inspectors can ask questions about track maintenance history, protocols and repair standards, and receive conversational answers.
“By being able to detect early defects in the rails, it saves not just money but also time – for both crew members and riders,” Demetrius Crichlow, president of New York City Transit, said in a press release. “This innovative program – which is the first of its kind – uses AI technology to not only make the ride smoother for customers but also make track inspector’s jobs safer by equipping them with more advanced tools.”
The MTA operates 36 subway lines, with 472 stations and 665 miles of track across Manhattan, Queens, Brooklyn and the Bronx. In 2023, the city’s subway system served approximately 3.6 million daily riders.
According to a January report from the transit agency, daily subway ridership in New York City increased by an average of 7.3%, or 214,000 riders, likely in response to the city’s newly implemented congestion pricing plan that started on Jan. 5, in which passenger vehicles, motorcycles, trucks and non-commuter buses entering the city’s Manhattan central business district, below 60th street, must pay a $9 toll fee.
