With their transit ride-share pilot, MARTA and Georgia Tech give passengers a ‘first-mile, last-mile’ solution
FILE: Georgia Tech and MARTA have partnered on a 6-month pilot of a new on-demand rideshare service, connecting riders in three zones across the region to MARTA bus and rail service. This pilot sought to test how on-demand shuttles can be used to make it easier and faster for riders to get to and from their destinations using MARTA.
When LaQuetta Ferrell learned about MARTA Reach, an on-demand pilot ride-share service in Atlanta, she eagerly started using the service that same day.
Ferrell’s commute to and from work had become a slog. She was getting up at 4:30 a.m. on weekdays to catch a bus and train to get to her job in downtown Atlanta by 7. She had to walk uphill and several blocks to the bus stop, wearing a brace for a worsening knee issue. Her knee hurt, and on days when the heat descended on Atlanta like a stifling blanket, she’d arrive home soaked in sweat. The one-way trip took 45 minutes on a good day but sometimes up to an hour and a half, versus the 15 minutes it would take to drive to work if Ferrell had a car.
MARTA Reach, a six-month pilot launched in March by the Metropolitan Atlanta Rapid Transit Authority (MARTA) in collaboration with the Georgia Institute of Technology, offered both convenience and a shorter commute. Each weekday morning, Ferrell called for a ride through an app on her phone, and a MARTA Reach shuttle picked her up across the street from her home and took her to a MARTA train station.
On the way home, she’d call for another ride from the train station and usually got picked up in less than five minutes, instead of waiting sometimes an hour for a bus.
“MARTA Reach really came in handy for me,” says Ferrell, an administrative assistant at the U.S. Environmental Protection Agency. “It’s great for me. It works well.”
MARTA and Georgia Tech launched the pilot to address what’s known as the “first-mile, last-mile” issue facing many residents like Ferrell, who don’t have easy access to a bus stop or train station. During the pilot, which ran from 6 a.m. to 7 p.m. Monday to Friday through August, users could call for a MARTA Reach ride by app or phone. A ride cost $2.50, the same as a regular MARTA fare, and transfers were free. All MARTA Reach vehicles, repurposed from the agency’s paratransit service for people with disabilities, are wheelchair accessible.
The pilot was initially launched in three zones and expanded in May to include several other neighborhoods. In collaboration with Georgia Tech, MARTA chose zones that were different from each other, seeking to determine how on-demand transportation would work in a residential area versus a mixed-commercial one or a more industrial location. Would riders be using the service mostly to get to work? To go shopping? Meet friends? The goal was twofold: to improve service for existing MARTA users and ideally, attract new users who might opt for transit over driving if it’s convenient enough.
Demand grew quickly, from fewer than 100 rides weekly when the pilot launched to more than 600 in early August. By late August, when the pilot ended, MARTA Reach had served more than 8,300 passengers and was projected to hit 1,250 rides weekly if the service had continued into September.
MARTA is now evaluating data from the program to understand how riders used it and determine, as the agency undertakes a redesign of its bus network, whether to extend the pilot or make the service permanent. Anthony Thomas, MARTA’s program manager for customer experience innovation, says preliminary data showed that many riders, like Ferrell, were using the service regularly.
“People are really excited about the program,” he says. “And we have been very excited about the uptake in service. We see lots of riders as well as very committed riders, individuals that were taking multiple rides a day, every day.”
And though MARTA Reach was designed to carry passengers relatively short distances, that convenience can make a profound difference in people’s daily lives, Thomas says.
“On paper it might look like, oh, that bus is pretty close. It’s only a 10-minute walk,” Thomas says. “But when you’re on the ground and it’s 95 degrees or you have groceries or you have kids with you or a stroller, that 10-minute walk becomes a barrier for folks, and they might just decide to hop into a car.
“For people who are on the lower-income spectrum, having to afford a car is a big burden. So being able to replace that trip with a $2.50 trip with MARTA is, I think, something that is extraordinarily powerful and very beneficial to the communities we operated the service in.”
The origins of MARTA Reach date back a decade, when Pascal Van Hentenryck, now a Georgia Tech professor of engineering and computer science, was leading a group of researchers in Australia focused on using data science to solve major challenges in areas including public transportation. Working in Canberra, Australia’s capital, the team at NICTA — Australia’s national information and communications technology research center — noticed taxis going back and forth from the airport to the Parliament building and many buses that were running empty or with few passengers.
The researchers envisioned a system that would remove some of the empty buses and instead use taxis to connect passengers with high-frequency buses. The group did some early planning, and when Van Hentenryck returned to the U.S. to work at the University of Michigan, he launched a ride-share pilot that offered free transportation on shuttle vans around campus and to several surrounding neighborhoods. Van Hentenryck and a team of students built an app for the service, which ran for four months in 2018.
“It was amazingly successful,” says Van Hentenryck, now the A. Russell Chandler III chair and a professor of industrial and systems engineering at Georgia Tech. “It was only running from 6 p.m. to midnight, but we had 400 students using the system every day.”
That success led Van Hentenryck to think bigger — about combining public transit with an on-demand ride-share service for Atlanta residents lacking access to transit. A sprawling metropolis with the ninth largest metro area population in the country, Atlanta has a network of MARTA bus routes linked to a rapid transit train system with 38 stations. But providing transit access for the region’s nearly 6.1 million residents is a pressing challenge as Atlanta continues to grow. Buses don’t serve the entire region, and some routes are underutilized.
Leaders at MARTA had been thinking about how to address the first-mile, last-mile issue when Van Hentenryck approached the agency in 2021 with a potential solution. He and his students had secured a $1 million grant from the National Science Foundation to develop an on-demand ride-share system. Drawing on the University of Michigan pilot, Van Hentenryck’s team would build out the technology and apps needed for the service and partly subsidize its operation.
“Clearly, it was a no-brainer at that point,” Thomas says. “It’s something that we’d wanted to test out in our system for a while. The stars had aligned, and it was a great opportunity to explore it at this time.”
Building a Solution
Van Hentenryck and a team of about seven Ph.D. students developed an Azure-based suite of technology for the pilot, including a routing system and separate apps for riders, drivers, and the administrative system.
The team started with the app developed for the Michigan pilot but building out technology designed for a college campus to serve a complex urban transit system spread out over a much larger geographic area proved challenging.
So the students worked closely with Microsoft to implement and optimize the app on Azure. They leveraged several key Azure capabilities to quickly build apps, improve data processing, and enhance the security of user data. Azure allowed the Ph.D. students to quickly set up and scale their app so they could focus on what mattered most — building a solution to increase mobility in metro Atlanta.
Azure enabled the automation of some tasks that would have been difficult to build from scratch, but since the system was new, there was no data to inform machine learning algorithms. The team quickly learned there were variables they hadn’t accounted for — in particular, driver behavior.
For one thing, MARTA Reach drivers drove more slowly — much more slowly — than Van Hentenryck had anticipated. That was great from a safety perspective, but it required the team to adjust the system accordingly. And in the early days of the pilot, drivers had few passengers and would sometimes not be paying attention to alerts about ride requests, so the team added functionality to quickly reallocate another vehicle when a driver was unresponsive.
“It’s human nature that if you’re sitting idle for 20 minutes, you’re going to zone out,” says Connor Riley, a former Georgia Tech student who worked on the pilot with Van Hentenryck and fellow Ph.D. student Anthony Trasatti and has since graduated.
“We had to do things to make sure that when a request came in, a driver had the information and was alerted to that request so that performance didn’t suffer,” Riley says.
As the pilot got underway and drivers got to know their regular passengers, another wrinkle developed. Drivers would sometimes drop passengers off at home or at non-designated stops, providing exemplary service while inadvertently mucking up the system.
“The drivers will go a long way to make sure the riders are happy. But at the same time, obviously that completely changed the optimization,” Van Hentenryck says. “At the end of the day, these are systems that are operated by people, and who are serving people in a human environment. And those factors are really difficult to predict.”
In response to feedback from passengers and MARTA, Van Hentenryck’s team added additional shuttle stops to the system and several new features, including a trip history so riders can easily repeat a route by clicking a button, and the ability to enter an address and find the nearest stop. The team also developed functionality for MARTA dispatchers to request rides for passengers who wanted to call rather than using the app.
“We wanted to make sure that people who either can’t afford a smartphone or maybe prefer not to use technology have access to the system as well,” Thomas says.
Over time, patterns emerged. Rising before 6 a.m. to monitor MARTA Reach rides real-time on a dashboard, the Georgia Tech team noticed many passengers traveling the same routes daily; Van Hentenryck estimates about 60% to 70% of MARTA Reach trips were commutes to work. Other regulars used the service for shopping, with Walmart and Kroger stores among the most popular destinations.
With MARTA Reach, “People don’t have to wait a long time for a bus and also don’t have to walk from the grocery store, carrying their packages to the bus stop,” says Hongzhao Guan, a member of the Georgia Tech team. “They could take their shopping cart to the parking lot, then take their bags and move right onto a shuttle. It’s very convenient.”
Transit agencies around the country have grappled with a shifting landscape impacted by competition from ride-share services such as Uber and Lyft, lower ridership during the Covid-19 pandemic, and labor shortages. Agencies in several other cities, including Los Angeles and Salt Lake City, have also launched on-demand ride-share services.
Van Hentenryck believes the MARTA Reach model could be replicated in other cities to connect riders with rapid transit bus or train service, in large part because it was achieved by just a handful of students working in cooperation with Microsoft. The application architecture is ready to scale to other metro systems if and when needed, without the need for a huge investment in staff or infrastructure.
“I think the biggest potential is going to be in mid-size cities, where you can connect people with a backbone of rapid transit using shuttles,” Van Hentenryck says. “I think that’s where the market is.”
For Guan, seeing how people used MARTA Reach, and the service the pilot provided, was gratifying.
“As Ph.D. students, normally we spend our days in front of a computer, running computational experiments and checking our results,” he says. “But MARTA Reach gave us an opportunity to test our idea in the real world. We received a lot of positive feedback from customers and see that they really rely on this service. I feel really proud that we helped local communities.”
Ferrell, for her part, hopes MARTA Reach will continue. She became friendly with her drivers, who would sometimes drop her inside her housing complex. She was an informal ambassador for MARTA Reach, putting flyers around the complex and in her office break room to let people know about it.
“I told a lot of people about it,” she says. “I love the service.”