Will Driverless Cars Increase Traffic Congestion?

Among the many things driverless cars will bring to society, a reduction in congestion is supposed to be one, due in large part because of their ability to travel much more closely together.  Except that future is by no means certain, as a recent paper from Arthur D. Little highlights.  It argues that, in the short-term at least, congestion is likely to get worse after the introduction of autonomous vehicles.

This is likely to occur as driverless cars share the roads with human drivers in a combination that the researchers believe will increase traffic jams by around 16%.  The results emerged after a micro-simulation was undertaken at a major intersection in Frankfurt that is notorious locally for the scale of traffic jams during the evening rush hour.

The five lane highway has two lanes turning left, with one turning right.  The speed limit is 80 km/hr and it is largely controlled by traffic lights.  The simulation was designed to measure the impact of three distinct scenarios on the flow of traffic:

  • A scenario where all cars were driven by humans
  • A scenario where all cars were autonomously driven (and traffic rules were adjusted accordingly)
  • A scenario with a 50/50 split of the two

Smooth flow

The simulation suggests that around 43 vehicles would pass through each green phase of the traffic light when the cars were all driven by humans.  This fell to 36 vehicles when there was a 50/50 split, with the fall due to the strict obeying of the rules by the autonomous vehicles that restricts risky behavior such as switching between lanes.

However, when the roads were fully autonomous, some 506 vehicles were able to pass through the light, due in part to optimized traffic rules, but also to the reduced 2.5m distance between each vehicle, platooning and a higher speed limit of 90 km/hr.  These are all things the researchers believe are feasible given the technological advances of driverless cars.

“Higher demand for mobility and greater urbanization are making congestion a growing issue within the world’s cities. Our micro-simulation found that while fully autonomous vehicles would solve this problem, the initial scenario we are likely to see, with a mix of robot and human drivers actually leads to greater traffic jams. Things would get worse before they get better. Clearly switching to 100% autonomous vehicles and changing traffic rules would require radical, disruptive action, posing serious questions for cities and society. Our work therefore aims to provide a starting point for further research and wider discussion, ahead of the future introduction of autonomous vehicles,” the researchers say.

Traffic improvements

This isn’t the first analysis to explore this issue of course. A recent paper from Rutgers University-Camden presents a slightly more optimistic picture in terms of the benefits smoother traffic flow can bring however.  It reminds us that we don’t need huge numbers of autonomous vehicles in traffic to derive benefits in terms of flow.

The researchers recently demonstrated their work to industry representatives, using virtual reality to highlight how a single autonomous vehicle can benefit at least 20 human-driven cars.  The research found that the autonomous vehicle was effective at controlling the flow of traffic by minimizing the stop-and-go waves that are so common when cars are driven by humans.  The team believe that if just 5% of cars are autonomous it could reduce total fuel consumption by up to 40%.

“Most of the policymakers, car manufacturers, car dealers, and others we talked with were very impressed with the research results and got a positive feeling about autonomous vehicles,” the team say. “They all agreed that the impact on real traffic economy and environmental impact could be of great importance.”

Suffice to say, these are all based on simulations at the moment, but it will be fascinating to see whether their predictions come to fruition when we do get driverless vehicles on the road.

Related

Facebooktwitterredditpinterestlinkedinmail