MIT And Stanford Show Robotaxis Could Replace Private Cars And Public Transit

As reported by Forbes: Someday, when fully autonomous vehicles are a reality, private car ownership and public transportation might plummet in favor of faster, better and cheaper mobility-on-demand services fulfilled by driverless cars. Imagine Uber powered by fleets of Google self-driving cars.

Such services underlie intriguing future scenarios where “robotaxis” provide door-to-door service while enabling significant reductions in transportation cost, enhancing mobility for millions saddled with limited access to private and public transportation, relieving congestion, and reducing the need for parking.

These scenarios are built on the fact that cars are relatively expensive but go mostly unused—cars are parked, on average, more than 90% of the time. Robotaxis would enable much higher utilization by sharing otherwise unused cars. This allows the purchase, maintenance and insurance cost to be spread across a large number of users on a pay-as-you-go basis, thereby increasing access and reducing cost for everyone. Also, since passengers don’t need to find parking, travel times, congestion, cost and space requirements would go down.

Robotaxis would be more efficient and convenient than emerging car sharing services like Zipcar and Car2Go in that the robotaxis go to the passenger, and return by themselves to the appropriate staging areas or to the next customer.

Robotaxis would be much less costly than traditional taxis and limousine services because there is no need for a human driver. (More on the jobs issue later.) Robotaxis would also deliver the added safety and performance benefits of self-driving cars.

There are, however, significant logistical and financial challenges to creating large-scale car-sharing services—even assuming that driverless technology works. The design challenges boil down to this: Can the fleet be sized and operated with acceptable service at a viable price?

For example, fielding too many cars to meet rush-hour demand results in high capital cost and a lot of idle cars during non-peak hours. Too few cars drive down service. Poor routing leads to a lot of empty miles, adding to cost, congestion and poor service. Even worse, routing algorithms must contend with stochastic demand that could easily lead to fleet imbalances and cause unpredictable and unacceptably long wait times.

A paper by a group of MIT and Stanford researchers reports several advances in addressing these challenges. The researchers developed rigorous methods to determine fleet sizing and manage robotaxi routing while ensuring attractive service levels. What’s more, they demonstrate that their methods work using actual traffic data and road networks of Singapore and New York.

In Singapore, the researchers applied their methods to extensive governmental data on travel patterns, traffic flows and road networks to simulate a large-scale robotaxi system. Rather than just replacing car traffic, the researchers show that a robotaxi system could handle all transportation needs—including private and public cars, taxis, scooters, buses, trains, etc.

Yes, you read that right. Their analysis showed that a fleet of 250,000 robotaxis could replace all modes of personal transportation and fulfill the transportation needs of the entire Singapore population. Maximum wait time with this fleet size is about 30 minutes during rush hours—and significantly lower during non-peak periods. Travel times would approximate current times.

Increasing the fleet to 300,000 vehicles brings maximum wait times down to less than 15 minutes. To put this number in context, there are about 800,000 total number of passenger vehicles in Singapore.

MIT Professor Emilio Frazzoli, one of the paper’s authors, shared this thought about his team’s work:

"Our study was more of a thought experiment: we assumed that there were no other means of transportation available. This is clearly unrealistic—but I think it sends a compelling message."

Compelling indeed: Theoretically, robotaxis could meet all of Singapore’s transportation needs at today’s service levels while eliminating 500,000 cars and all buses and trains.

Others are also studying the potential of robotaxi-enabled car sharing. Some studies, in fact, conclude that robotaxis could replace an even larger percentage of human-driven cars. This study is among the first to use actual transportation data of such scale.

Larry Burns, Professor of Engineering Practice, University of Michigan and former corporate vice president in charge of research, development and planning at General Motors, praised the research. Burns, along with several colleagues, reached similar results in earlier research conducted at Columbia’s Earth Institute. That study, however, used simpler analytics and simulation models rather than actual data.

“Every city is different in terms of road networks, traffic flows, trip densities, and congestion,” Burns told me. “MIT’s work very good, and is the right thing to do if you are planning for a specific system.”

Some might recoil at the prospect of robotaxis taking riders away from public buses and trains. Brad Templeton, who coined the “robotaxi” term and is chair of Computing & Networks at Singularity University, argues, however, that robotaxis would not only be more convenient than public transit but also more environmentally advantageous. In several articles, including The Decline of Mass Transit and Is Green U.S. Mass Transit a Big Myth?, Templeton argues that well designed robotaxis will beat the energy efficiency of public transit by large margins. Templeton doesn't argue for replacing existing transit systems but rather that robotaxis might well diminish the need for major extensions and new systems.

Given the politics of public transit funding, however, a more likely robotaxi adoption strategy is to target the displacement of cars and taxis. In Singapore, which has a very sophisticated public transportation system, cars and taxis account for about 33% of total trips. Might a robotaxi fleet of less than 100,000 cars eliminate the need for all 800,000 privately owned, human-driven cars and taxis?

A glimpse of this future shown in a related paper, where several of the same researchers applied similar methods to model a system that could handle all New York City taxi traffic. In that study, the researchers showed that their routing algorithms could serve the same demand with a 40% reduction in fleet size. The savings result from intelligent coordination of the robotaxis to minimize congestion, keep the system in balance and better serve anticipated demand.

The research also shows that robotaxis are financially viable.

In New York, the economics for replacing taxis is straightforward. Reducing the cost of drivers—on top of a 40% reduction in vehicles—leaves a wide margin for a sustainable robotaxi business model.

For Singapore, the researchers estimated that direct cost per mile for the robotaxis would be about 30% less than human-driven cars. This analysis was based on conservative assumptions about technology cost and actual operational data from current car-sharing services, like ZipCar. If the value of the time saved is considered, the savings increased to almost 50%. Again, such levels of cost reductions leave ample room for a sustainable business.

In addition to the tremendous cost and time savings, another major benefit of robotaxis is increased mobility at a practical cost for the disadvantaged, disabled and elderly with limited access to cars or unable to drive. A US Bureau of Transportation Statistics survey found that almost 15 million people, six million of whom are disabled, have difficulties getting the transportation they need. This number will rise. The Los Angeles Times reports that by 2030, up to a quarter of the nation’s licensed drivers will be older than 85. Not having easy, affordable transportation or losing the ability to drive altogether has been linked to lower employment, increase in depressive symptoms and a decline in out-of-home activity levels.
Massive disruptions would come along with the benefits, however.

There is much to learn, for example, about the secondary effects of making car travel cheaper and more convenient. Will this shift usage from public transit? Will it drive up overall demand and increase pollution and congestion? Will it enhance urban, suburban and exurb sprawl?

The impacts on jobs and profits will be substantial. More than $2T is spent each year in the US on car-related spending, encompassing suppliers, carmakers, dealers, financing, service, repairs, insurance, energy, rentals, taxes, etc. As I've previously discussed, massive car sharing has the potential of eliminating or redistributing a significant portion of these revenues through new business models and changes in the competitive landscape.

Professional drivers, for example, could suffer huge job losses. In New York City alone, there are over 50,000 licensed taxi drivers and about another 50,000 other professional drivers of black cars, livery services and other For-Hire Vehicles (Source: 2014 NYC Taxicab Fact Book). (I explored this issue in depth in several previous articles, including one entitled Will The Google Car Force A Choice Between Lives And Jobs?)

The disruptions would reach far beyond professional drivers. Take automakers and car dealers, for example.

Car dealers could be disintermediated if robotaxis are sold as large fleets to robotaxi operators, rather than to through dealers to individual owners. Car dealers in the US handle more than $650 billion in new and used cars sales today and, as Tesla is finding out, jealously guard their position in the automotive value chain.

An increase in fleet sales and in total mile travel due to cheaper transportation would be good for automakers but robotaxis could hurt them in other ways.  Profits would be squeezed if robotaxis cut into the sales of large expensive models that provide most of today’s margin. That’s because most car trips involve only one or two persons. So, if robotaxis allow riders to call for the type of car needed, when they need it, customers might opt for smaller, less expensive cars. Gone might be the days when buyers choose minivans just to accommodate the occasional family outing or car-pooled soccer game. They might opt buy the smaller car, and request larger robotaxis when they need it. The same rationale might diminish the tendency to purchase second or third cars for occasional use.

A lot of invention and innovation is needed before a robotaxi pulls up to your door and, in doing so, induce broad social and economic disruptions. Estimates of when this might happen range from a few years to never. Many hard issues remain to be solved—but many forces are working towards making that eventuality come sooner rather than later.

Fasten your seat belts; we are in for a wild and bumpy ride.