Traffic Sensor Locations Problem For Estimation Of Stochastic Origin-Destination Traffic Demands Matrix

Origin-destination(OD) traffic demands estimation by traffic counts is becoming an increasingly important issue in the field of transportation research. The corresponding sensor location problem for estimating OD matrices is getting more and more attention also. Conventionally, the OD matrices estimation from traffic counts problem usually considers how to accurately estimate the mean of the OD demands while ignoring its statistical property. Actually, the OD demands for a typical period(e.g. morning peak) of different OD pairs are not deterministic and fluctuate from day to day. It is not enough to show the actual OD demands by the mean of OD demands only. It is essential to investigate the correlation between two arbitrary OD demands. Based on the previous research this thesis mainly investigates the sensor location model for estimating both the mean and covariance of OD demands under network uncertainty.Chapter 1 provides a brief introduction to the background, importance and literature review on the concerned problem. Chapter 2 introduces the preliminaries of this thesis. In Chapter 3 the model and corresponding solution algorithm of the stochastic OD demands estimation problem is proposed. In Chapter 4, traffic sensor locations problem for estimation of stochastic OD demands is fumulated as an optimization model based on the model of Chapter 3. A chance constaint is proposed in this model to fully consider the stochasticity of OD demands while optimizting the sensor locations. In Chapter 5, the numerical experiments are provided to show the properties of the proposed traffic sensor locations problem and the convergence of the proposed genetic algorithm. It is found that the chance constraint can save the compoutational efforts while grarenteeing the computational accuracy. The conclusions and further studies were shown in in Chapter 6.