Machine-Learned Prediction Equilibrium for Dynamic Traffic Assignment

We study a dynamic traffic assignment model, where agents base their instantaneous routing decisions on real-time delay predictions. formulate mathematically concise model and derive properties of the predictors that ensure prediction equilibrium exists. demonstrate versatility our framework by showing it subsumes well-known full information models, in addition to admitting further realistic as special cases. complement theoretical analysis an experimental study, which we systematically compare induced average travel times different predictors, including machine-learning trained data gained from previously computed flows, both synthetic real road network.