Convex relaxation problems in traffic control

Gabriel Gomes
University of California, Berkeley (UC Berkeley)

Traffic systems are often represented with network flow models. Traffic controllers are then tuned by solving optimization problems on those networks. In this talk I will present several instances of this: the ramp metering problem and a few versions of the offset selection problem for signalized intersections. These will be formulated as non-convex and non-differentiable optimization problems. Interestingly however, we will find optimal solutions to all of them using relaxation techniques.


Back to Workshop III: Traffic Control