MIT's Transit Lab, a research group that has worked with public transport operators from Boston to Hong Kong for decades, is building an artificial intelligence platform to help transit agencies run their networks day to day. The lab said on 30 September that it had won $2.1 million from Google.org to develop the Public Transit Intelligence Hub, or PTIQ, an open-source system that will bring an agency's real-time monitoring, operations control and passenger information into a single tool.
PTIQ will combine predictive models, optimisation software and reasoning based on large language models, and present the results through a decision-support screen for the staff who manage buses and trains from an agency's control centre. The aim is to help them respond faster when something goes wrong, and to give riders better information when it does.
"Our goal isn't to automate those decisions, but to make sure people have best information," said Awad Abdelhalim, associate director of the Transit Lab, who leads the project as co-principal investigator and technical lead.
What PTIQ Is Meant To Fix
The project targets a problem familiar to anyone who has worked in a transit control room. Agencies run separate systems for vehicle tracking, operations, staffing and customer messaging, often bought from different vendors over many years, and controllers must switch between screens to piece together what is happening when a train breaks down or a bus route falls behind schedule.
Jim Aloisi, the MIT lecturer and former Massachusetts secretary of transportation who manages the programme, said PTIQ will "connect siloed environments," providing benefits for both the workforce and riders. A controller facing a disruption could see the likely knock-on delays, compare options such as short-turning trains or pulling in spare buses, and send consistent updates to passengers, without stitching the information together by hand.
Built For Agencies, Not Just One City
Because PTIQ will be open source, agencies of any size should be able to adapt it rather than buy a proprietary platform. That matters for the many mid-sized and smaller operators that lack the budget or in-house data teams of New York's MTA or Transport for London.
The three-year project brings together the MIT Transit Lab, the MIT Mobility Initiative and Northeastern University, with Northeastern professor Haris Koutsopoulos leading the work through a transit research consortium. Google.org will add pro bono help from its engineers and AI product teams.
The Hard Part Is Not The Technology
Jinhua Zhao, who heads MIT's Department of Urban Studies and Planning and founded the MIT Mobility Initiative, was frank about where the project could stumble. "The hard part of integrating AI in transit is not the technology; it's the institution," he said.
Transit agencies are public bodies with union agreements, procurement rules, safety regulators and long-serving staff who know their networks intimately. A tool that recommends actions in a control room has to earn the trust of those staff, fit into established procedures and stand up to scrutiny when its advice turns out wrong, which explains the project's insistence that people keep the final say.
The Transit Lab brings an unusual track record to that challenge. It has worked with agencies including the Chicago Transit Authority, Transport for London, Hong Kong's MTR and Boston's MBTA, and has spent years studying smart-card fare data to model how passengers move, along with operational problems such as unplanned bus driver absences.
Part Of A Wider Google.org Bet On Public-Sector AI
The grant comes from the Google.org Impact Challenge: AI for Government Innovation, which named 15 recipients in September to share $30 million. The other projects range from Code for America's work on free tax filing to Johns Hopkins University's road-safety tools for city transport departments and a University of Cincinnati project to predict infrastructure failures, and each team will build open-source tools that other governments can reuse.
Google has a growing stake in transit technology. Its public sector arm worked with New York's MTA on TrackInspect, a pilot that mounted Google Pixel phones on subway cars to pick up vibrations and sounds from the rails, and the system identified 92% of defect locations that inspectors later confirmed.
A Market Already Moving
PTIQ will not enter an empty field. Commercial vendors already sell AI to transit agencies, and Optibus, the Israeli scheduling software company, launched an AI agent for transit operations in June that builds schedules around labour agreements, finds qualified drivers for open shifts and locates vehicles during disruptions. "AI is a co-pilot, not the captain," said Amos Haggiag, the company's chief executive.
Where PTIQ differs is in its open-source model and its focus on pulling an agency's existing systems together rather than replacing them. If it works, agencies could use it alongside commercial tools, and vendors could build on it.
Why Agencies Need Better Tools Now
The timing suits an industry still rebuilding after the pandemic. US transit riders took 8.1 billion trips in 2025, up 6% on the year before but still only about 81% of pre-pandemic levels, according to the American Public Transportation Association, and many agencies face tight budgets as federal relief funds run out.
Reliability is one of the few levers agencies control that can win riders back, and a control room that reacts faster to disruptions can keep more trips on time. Better passenger information also matters, because riders forgive delays more readily when they know what is happening and what to do next.
The Data Is Already There
Transit networks generate huge amounts of data from GPS units on buses, passenger counters, fare gates, signalling systems and, increasingly, sensors mounted on vehicles and track. Much of it sits in separate databases and is used after the fact for planning rather than in real time for operations. Cities are adding more connected infrastructure on top, from smart streetlights to traffic sensors, a trend we followed as cellular moves up the lamppost, and transit agencies stand to gain from tools that can make sense of it while services are running.
What To Watch As PTIQ Takes Shape
MIT has not yet named the agencies that will pilot PTIQ, and the first test will be which operators sign up and how their control-room staff respond. Agencies will also want to see how the system handles the reliability demands of live operations, how it protects passenger data, and how its large language model components avoid giving confident but wrong advice during an incident.
For technology suppliers to the transit sector, an open-source hub backed by MIT and Google could become a reference point that shapes how agencies think about data integration, and vendors may need to show their products work with it.
A Control Room That Sees The Whole Network
PTIQ is a modest grant by technology standards, but it targets a stubborn gap in how cities run their transport. Transit agencies have plenty of data and a growing range of AI tools, yet the people managing services minute by minute still work across disconnected systems, and that is where delays turn into missed connections and frustrated riders.
If MIT can deliver an open tool that agencies trust in the control room, it could spread far beyond the cities where it is first tested. The project's success will rest less on the sophistication of its models than on whether controllers find its advice useful on a bad morning, which is exactly the institutional test Zhao warned about.
