How automated vehicles could eliminate ghost traffic jams
Redmer de Haan defended his PhD thesis at the Department of Mechanical Engineering on June 3rd.
In busy traffic, even a minor speed adjustment by one driver can trigger a chain reaction. The following driver brakes slightly more, the next driver reacts even more strongly, and before long a wave of slowing traffic moves backward through the road network.
These ‘ghost traffic jams’ reduce road capacity, increase travel times, and force vehicles to repeatedly accelerate and brake, which wastes energy and increases emissions.
The promise of vehicle automation
Automated vehicles have the potential to make better use of available road space by following each other more closely and more consistently than human drivers.
However, achieving this in real-world traffic is not straightforward. Vehicles are not identical. A fully loaded truck responds differently from a sports car, and even two similar vehicles may behave differently depending on their driver-assistance systems or operating conditions.
Ignoring these differences can actually make traffic instability worse.
A decentralized approach
The research of Redmer de Haan focuses on a practical control strategy in which each vehicle only uses information from the vehicle directly ahead.
This decentralized approach avoids the need for central coordination or detailed knowledge of every vehicle in a traffic stream. As a result, it remains scalable and practical for long strings of vehicles on public roads.
Despite relying on limited information, the approach can still contribute to smoother and more stable traffic flow.
The challenge of delays
One of the biggest obstacles is delay. When a vehicle receives a command to accelerate or brake, its motion does not change instantly. These delays can significantly affect how closely vehicles can safely follow each other.
The dissertation addresses this challenge in three complementary ways. First, it develops control algorithms that remain effective despite delays. Second, it uses prediction techniques to compensate for delayed vehicle responses. Third, it adapts the desired following behavior to explicitly account for delays. Together, these methods help maintain stable vehicle behavior even when response times are not instantaneous.
Testing the theory on real vehicles
The proposed algorithms were not only studied in simulations but were also tested on real vehicles with different driving characteristics and response delays.
The experiments confirmed an important finding: control strategies that ignore differences between vehicles tend to amplify speed variations as they travel through a line of vehicles.
The newly developed algorithms, however, successfully prevented this amplification while allowing vehicles to follow each other closely.
Driving closer without creating traffic waves
A key result of the research is that automated vehicles were able to comfortably maintain a following time gap of just 0.5 seconds.
For comparison, human drivers are commonly advised to keep a gap of around two seconds.
This means that future automated traffic could use road capacity much more efficiently while still maintaining stable traffic flow.
A step toward smoother and more efficient roads
The central conclusion of the dissertation is clear: vehicle automation must account for differences between vehicles, delays, and limited information.
When these factors are incorporated into the design, automated vehicles can drive closely together without amplifying speed fluctuations. If they are ignored, automated vehicles may still create or worsen traffic waves.
By addressing these challenges, the research takes an important step toward reducing congestion, making better use of road infrastructure, and lowering unnecessary energy consumption on public roads.
Title of PhD thesis: . Supervisors: Dr. Erjen Lefeber, Dr. Igo Besselink, and Dr. Tom van der Sande.