Driverless taxis are moving from pilot projects to real-world deployment. Researchers at EPFL are studying what cities must do it they want to integrate robotaxis without increasing congestion or weakening public transport.
Switzerland recently launched its first driverless taxi service in Furttal, near Zurich. As the so-called robotaxi services become a reality, researchers are increasingly focusing not just on their technical feasibility, but on how large fleets of autonomous vehicles can be integrated into transport systems, while also extending mobility options to currently underserved rural regions.
Kenan Zhang, professor at the EPFL Laboratory for Human-Oriented Mobility Eco-System (HOMES), studies how robotaxis can be integrated into transportation networks. She analyzes many factors, from fleet allocation to parking placement and social reception. "Robotaxis are technically feasible, but their future depends much more on regulation, public acceptance, economics, and integration with cities and public transport systems," says Zhang.
In a series of studies published in Transportation Research, Zhang and her team present artificial intelligence methods that are able to coordinate large fleets of autonomous vehicles and optimize where those vehicles wait and park when not carrying passengers. The findings suggest that future robotaxi networks could be coordinated in ways that reduce waiting times, improve vehicle utilization, and expand service availability, including areas that are traditionally underserved by on-demand transport.
Serving remote areas
Certainly, taxi drivers naturally circulate around busy areas such as urban centers, airports, and train stations, leaving suburban and less densely populated areas with fewer transport options. A fleet of autonomous vehicles, on the other hand, could tackle this limitation by dispatching vehicles strategically, helping maintain service in suburban areas.
To this end, cities may need to coordinate thousands of autonomous vehicles, ensuring that enough vehicles are available where passengers need them. To study a centrally coordinated fleet management, Zhang and her colleagues developed reinforcement learning algorithms using real-world data from the city of Shenzhen, in China. In this centralized approach, rather than allowing each vehicle to make independent decisions, the system continuously redistributes vehicles across the city according to anticipated demand, helping maintain service quality and ensuring availability in neighborhoods that are often overlooked by conventional ride-hailing operators, such as Uber.
However, robotaxis could solve one urban headache while creating another. "As autonomous vehicles reduce the need for traditional parking, they may also spend more time circulating on the streets while waiting for passengers," says Zhang. To address this issue, she suggests the introduction of new regulations, such as limiting driverless vehicles without passengers, charging fees for unnecessary circulation, or creating dedicated waiting areas.
Complementing public transport
Even if the technology is ready, one of the key factors for its successful deployment will depend on whether autonomous transport can complement, rather than undermine, sustainable mobility systems. In such a system, autonomous vehicles would become one component of a broader transport ecosystem, extending the reach of public transport instead of drawing passengers away from it.
However, the long-term cost of large-scale robotaxi services remains uncertain. If robotaxis become cheaper, more people may be willing to use them, increasing both their potential benefits and the risk of competition with public transport. "That would not be a good thing, because public transportation will always be more efficient in transporting travelers with limited vehicle space, road capacity, and energy consumption," says Zhang.
Switzerland represents a particular case when it comes to adopting autonomous taxi services. The country already has a high-quality public transport network with extensive coverage across most urban areas. These factors can limit the demand for robotaxis, in particular in the city centres.
A cascading effect
"Considering all these technical and economical dimensions, at the beginning adoption of autonomous taxis will not be evenly spread," believes Zhang. Major cities in countries such as the United States and China could undergo widespread deployment within the next decade. Then autonomous driving and robotaxis might benefit from a cascading adoption effect. "Once a city demonstrates that robotaxis can operate safely and efficiently, other cities with similar transport needs are likely to follow," concludes Zhang.
The next challenge researchers need to address is ensuring that the computational models developed remain effective when deployed in real cities, where travel demand, infrastructure, and human behavior are far more complex than in simulations.
