Vanderbilt: Improving public transit system with Google Cloud's AI

Using AI to improve public transportation

Public transportation infrastructure is an essential component in cultivating equitable communities. However, public transit agencies have historically struggled to efficiently serve their wide range of users. Smart sensors embedded in vehicles and roads can now provide a new and expanded understanding of how people use transportation in their daily lives. But in order to produce actionable insights, all of this collected data must be analyzed—accurately, promptly, and at scale.

Abhishek Dubey, Associate Professor of Electrical Engineering and Computer Science and director of the ScopeLab in the Institute for Software Integrated Systems at Vanderbilt University, sees this moment as an opportunity to use machine learning to make the public transit sector more efficient. He leads SmartTransit, a research team that collaborates with Chattanooga Area Regional Transportation Authority (CARTA) and Nashville WeGo to design efficient transit operation algorithms by using artificial intelligence and real-time data analysis at scale. Their goals are to reduce the costs of energy usage, improve customer experience, and lower the greenhouse emissions that affect community health and climate change. In 2022 Dubey joined Google Cloud Research Innovators program, which offers pioneering researchers across the globe access to Google experts, technical resources, additional Google Cloud credits, speaking opportunities, and training and support from mentors. With those resources, the team was able to jumpstart their pilot operation with CARTA.

Managing massive amounts of data from multiple sources

Working with CARTA, the team has already obtained around 32.3 million data points for electric buses and 29.8 million data points for diesel buses. They also collect weather data from multiple weather stations in Chattanooga at five-minute intervals. This data includes real-time temperature, humidity, air pressure, wind speed, wind direction, and precipitation. In addition, they collect traffic data, in the form of timestamped speed recordings from selected roads, at one-minute intervals.

Combined with Geographic Information System (GIS) data for real-time road conditions, this multimodal dataset is comprehensive but difficult to store, process, and analyze. Dubey explains that "the first challenge is storing the high-velocity, high-volume data streams. The second challenge is that the data is highly unstructured and irregular. Different data streams have to be synchronized and joined efficiently in order to run downstream applications such as monitoring systems, visualization dashboards, and energy and ridership prediction models."

With grants from the National Science Foundation's Smart and Connected Communities initiative (SCC) and the Department of Energy, Dubey and his team are piloting a project with CARTA to store these data streams on Google Cloud so they can be more easily analyzed to optimize efficiencies. The prototype includes an energy visualization dashboard to monitor the performance of the CARTA fleets over time. This dashboard is publicly available and allows anyone to query energy use and other statistical visualizations over a map of Chattanooga. "Once you deploy these tools and datasets on the cloud," Dubey says, "it is much easier to use, even for people who aren't experts on machine learning. Google's BigQuery and App Engine have been very, very useful and make Google Cloud much easier to use than other cloud vendors."

Transforming the public sector with data science

With easy access for offline model training and real-time updates, the prototype has already modeled $150,000 in fuel cost savings for CARTA–and using less fuel lowers greenhouse emissions that benefit whole communities. For another project, the team developed software to schedule paratransit and microtransit services on demand. In a pilot, results showed greater than 40% possible improvement in efficiency compared to baseline operations.

"I believe that the integration of data-driven methods with better operational research methods combined with a socially engaged design will have a broad impact," says Dubey. "Public transportation is a critical service for almost everybody that uses it. Applying data science at scale to cities has so much potential. Cities have data across all the departments; they just need the right set of computation resources, and familiarity with processing data at that scale. With these kinds of partnerships and access to the city data, we can transform the public sector."

Read more about Dubey's research. Learn more about the current Google Cloud Research Innovators' cohort. To find out how you can get started with generative AI for research and transportation, download the new 10-step public sector guide. With domain-specific use cases and customer stories from the city of Memphis, the state of Minnesota, the U.S. Department of Defense, and more, it offers a comprehensive guide to kickstart your gen AI journey.