The Metropolitan Transportation Authority (MTA) in New York City has partnered with Google for a groundbreaking pilot program focused on enhancing the reliability of its old subway network. Utilizing Google’s mobile technology, the effort aims to detect and resolve rail problems before they cause service interruptions. Named “TrackInspect,” the project signifies a considerable advancement in applying artificial intelligence and contemporary technology to public transportation.
Beginning in September 2024 and wrapping up in January 2025, the pilot project involved equipping certain subway cars with Google Pixel smartphones. These phones were responsible for gathering sound and vibration information to identify possible track issues. This data was subsequently evaluated by Google’s AI systems in the cloud, which identified zones that needed further examination by MTA staff.
“By spotting initial indicators of track deterioration, we not only cut down on maintenance expenses but also lessen inconveniences for passengers,” stated Demetrius Crichlow, president of New York City Transit, in an announcement made public in late February.
The collaboration between the MTA and Google forms a component of a larger initiative to update New York City’s 120-year-old subway system, which still confronts issues due to its outdated infrastructure and regular delays. Although the pilot program yielded encouraging outcomes, doubts persist about the potential expansion of TrackInspect, considering the financial limitations the MTA is experiencing.
Addressing delays through AI and smartphones
New York City’s commuters frequently encounter subway delays as a recurring issue. Towards the end of 2024, the MTA disclosed that tens of thousands of delays were occurring monthly, with December alone surpassing 40,000 incidents. These interruptions stem from multiple causes, such as track problems, construction activities, and crew shortages.
El programa TrackInspect se centra en abordar un aspecto crucial del problema: detectar y solucionar problemas mecánicos antes de que se agraven. Durante la prueba piloto, se instalaron seis teléfonos Google Pixel en cuatro vagones R46 del metro, reconocidos por sus asientos de color naranja y amarillo. Los dispositivos registraron 335 millones de lecturas de sensores, más de un millón de datos de GPS y 1,200 horas de audio.
Los teléfonos inteligentes se colocaron estratégicamente tanto dentro como debajo de los vagones del metro. Los dispositivos externos estaban equipados con micrófonos para captar sonidos y vibraciones, mientras que los internos tenían los micrófonos desactivados para evitar grabar conversaciones de los pasajeros. En cambio, estos dispositivos se concentraban únicamente en las vibraciones para identificar anomalías en las vías.
Rob Sarno, un asistente del jefe de vías de la MTA, desempeñó un papel crucial en el proyecto. Sus tareas incluían examinar los fragmentos de audio señalados por el sistema de inteligencia artificial para detectar posibles problemas en las vías. “El sistema destacó áreas con niveles de decibelios anormales, lo que podría sugerir uniones sueltas, rieles dañados, u otros defectos,” explicó Sarno.
La línea de tren A, seleccionada para el piloto, presentó un entorno de prueba variado con vías tanto subterráneas como elevadas. Además, incluyó segmentos de infraestructura recientemente construida, ofreciendo un punto de referencia para comparaciones. Aunque no todos los retrasos en la línea A se deben a problemas mecánicos, los datos recopilados durante el programa piloto podrían contribuir a resolver problemas recurrentes y mejorar el servicio en general.
Encouraging outcomes, yet challenges persist
The TrackInspect initiative produced promising results, as the AI system accurately identified 92% of defect locations that were confirmed by MTA inspectors. Sarno estimated his own accuracy rate in anticipating track defects from audio data to be approximately 80%.
The TrackInspect program yielded encouraging results, with the AI system successfully identifying 92% of defect locations verified by MTA inspectors. Sarno estimated his personal success rate in predicting track defects based on audio data at around 80%.
The program also included an AI-powered tool based on Google’s Gemini model, which allowed inspectors to ask questions about maintenance protocols and repair history. This conversational AI provided inspectors with clear, actionable insights, further streamlining the maintenance process.
Despite its success, the pilot program raises questions about scalability and cost. The MTA has not disclosed how much it would cost to implement TrackInspect across its entire subway system, which includes 472 stations and serves over one billion riders annually. The agency is already grappling with financial challenges, needing billions of dollars to complete existing infrastructure projects.
An increasing movement in transit advancements
A growing trend in transit innovation
Google has previously worked with other transportation agencies. The tech company has created tools to optimize Amtrak’s scheduling and has teamed up with parking technology providers to incorporate street parking information into Google Maps. Nonetheless, the size and intricacy of New York’s subway system make this project especially ambitious.
La red de metro de la MTA es la más grande de Estados Unidos, brindando servicio las 24 horas en muchas de sus líneas. Este funcionamiento continuo añade otra capa de complejidad a los esfuerzos de mantenimiento, ya que las reparaciones y mejoras a menudo deben realizarse junto al servicio activo. Con el uso de tecnología de inteligencia artificial y teléfonos inteligentes, el programa TrackInspect podría ayudar a la MTA a enfrentar estos desafíos de manera más eficiente.
Looking forward
Aunque el piloto de TrackInspect ha concluido, la MTA está investigando asociaciones con otros proveedores de tecnología para seguir mejorando sus procesos de mantenimiento. La agencia también está evaluando los datos del piloto para determinar su impacto en la reducción de retrasos y mejora del servicio. Las primeras señales sugieren que ciertos tipos de retrasos, como los causados por problemas de frenado y defectos en las vías, disminuyeron en la línea A durante el periodo del piloto. No obstante, la MTA advierte que se requiere un análisis más detallado para confirmar un vínculo directo con el programa.
While the TrackInspect pilot has ended, the MTA is exploring partnerships with other technology providers to further enhance its maintenance processes. The agency is also analyzing data from the pilot to determine its impact on reducing delays and improving service. Early indications suggest that certain types of delays, such as those caused by braking issues and track defects, decreased on the A line during the pilot period. However, the MTA cautions that further analysis is needed to confirm a direct link to the program.
For now, the pilot represents a promising step toward modernizing the MTA’s operations and addressing the challenges of an aging transit system. By combining the expertise of tech companies like Google with the experience of transit professionals, New York City may be able to deliver a more reliable subway experience for its millions of daily riders.
As Sarno reflects on the project, he emphasizes the potential of AI-driven solutions to transform public transportation. “This technology allows us to detect problems earlier, respond faster, and ultimately provide better service to our customers,” he said.
The MTA’s collaboration with Google underscores the potential of public-private partnerships to drive innovation in critical infrastructure. Whether TrackInspect becomes a permanent fixture in New York’s subway system remains to be seen, but its success highlights the possibilities of integrating cutting-edge technology into the daily lives of commuters.
