Putting pocket tech to work: NYC Subway deploys Google Pixels to detect track defects

The New York City subway system’s 665 miles of track require constant inspection to maintain safety and reliability. In a pioneering experiment, the Metropolitan Transportation Authority (MTA) partnered with Google to test whether everyday smartphones could revolutionize how track defects are detected.

Project overview: The MTA and Google’s four-month experiment utilized six Google Pixel phones mounted on A-line subway cars to collect comprehensive track condition data.

• The phones leveraged built-in sensors including accelerometers, magnetometers, gyroscopes, and microphones to gather information about track conditions
• The experimental system, dubbed TrackInspect, employed artificial intelligence to analyze audio, vibration, and location data
• The project processed 335 million sensor readings and 1,200 hours of audio, combining this with existing defect records

Technical performance: The smartphone-based detection system demonstrated remarkable accuracy in identifying track issues during the trial period.

• The AI-powered system successfully identified 92% of defects that were later confirmed by human inspectors
• MTA employee Robert Sarno played a crucial role by labeling audio data to train the AI models
• The technology proved capable of detecting problems using consumer-grade hardware that costs a fraction of specialized inspection equipment

Current inspection methods: Traditional track inspection relies on a combination of human expertise and specialized equipment deployed at regular intervals.

• Human inspectors physically walk the hundreds of miles of track to identify potential issues
• Specialized “train geometry cars” conduct detailed track examinations three times annually
• These conventional methods, while thorough, are time-intensive and may miss developing problems between inspections

Implementation roadmap: The successful experiment has led to plans for expanding the technology’s use within the MTA system.

• A full pilot project is being developed to integrate the technology into track inspectors’ daily workflows
• The system is designed to complement, not replace, human inspectors who remain required by regulations
• The goal is to enhance efficiency by helping inspectors identify and address track issues before they impact service

Looking ahead – practical implications: While the experiment shows promise for modernizing track maintenance, the technology represents an augmentation rather than a replacement of existing inspection protocols.

• The low cost and accessibility of smartphone-based detection systems could make similar programs feasible for transit systems worldwide
• Human expertise remains crucial for validating AI findings and performing repairs, suggesting a future where technology and human judgment work in tandem
• The success of this pilot program may encourage other transit authorities to explore similar technological solutions for infrastructure maintenance