Gerencsér, Balázs RI / Lovas, Attila BME
Non-linear state-estimation based traffic modelling relying on sparse measurement data
We present our approach to the theoretical and computational challenge of traffic data extrapolation, to fit a model via its dynamical system approximations to the real world measurement data available and get an overall picture of ongoing traffic. There is a highly non-trivial tuning for acceptable performance and accuracy given the high dimension of traffic state. The data source provided by the cooperation through the city administration has significantly lower dimension than the parameter space of the model we build, so various dimension reduction, regularization measures have to be taken to get a realistic outcome. The current mini-project is towards a broader perspective of research, where the vision is to get a high resolution estimate of traffic related emission, for which an intermediate milestone is getting a fine detailed insight on the traffic actually causing the emission. Consequently, we target an online, fast-responding tool that can be later used for smart city applications.
Online location: https://meet.google.com/ckd-siwc-nvx?authuser=0