Modeling stochastic human-driver car following behavior in oscillatory traffic conditions

Principal Investigator(s):

Raphael Stern, Assistant Professor, Civil, Environmental and Geo-Engineering

Project summary:

To calibrate stochastic ordinary differential equations for car-following behavior, researchers used existing vehicle-level trajectory data to calibrate models that are able to describe the stochastic nature of human driving. These models were calibrated using different criteria such as mean squared error in speed, spacing, and headway as the objective function in a gradient-based method. This allowed for different parameter optimization techniques to be compared. The resulting models were then compared with traffic flow simulations where some of the vehicles in the flow were replaced with autonomous vehicles. This provided insight into how we can expect traffic flow to change in the future when autonomous vehicles are present.

Project details: