Dynamic Original-Destination Matrix Estimation Method Research

Nowadays, traffic congestion, traffic accident, environment pollution has become serious problems in many cities. We need not only built more and more new roads, but also to enhance traffic control level. We should control traffic intelligently.Intelligent Transportation Systems(ITS) is the indication that transportation has entered the information era, which has been paid much attention in China and other developed countries. Now, the research on ITS has just been initiated in China, and lots of problem need to be studied. Dynamic origin-Destination(OD) matrix estimation is one of the key problems.Dynamic OD matrix describes the time variable traffic demand, which is the basic input of dynamic traffic assignment(DTA) models and many applicable microscopic traffic simulators. Generally speaking, it is difficult to obtain the dynamic OD matrix through traffic investigations. Therefore, the OD estimation algorithms in this thesis all are estimating based on link traffic counts. In this thesis, we not only introduce a dynamic OD estimation algorithm based on conventional Kalman filtering, but also bring in a new kind of dynamic OD estimation algorithm based on Inverse-of-matrix-free Kalman filtering. The algorithm is validated by simulation results.The main works and contributions to state of the art of this thesis are as follows:(1) Bring in a new kind of dynamic OD estimation algorithm based on Inverse-of-matrix-free Kalman filtering for the first time. Similar to the conventional Kalman filtering, the state variables are defined by deviations of OD deviations from historical estimates, which can easily incorporate the structural information about travel modes in historical data. Additionally, the new algorithm guarantee the matrix P's Symmetric nature definite, and avoid filtering Divergent. The new kind of algorithm doesn't need inversing of matrix, so it's a simple algorithm.(2) For the key assignment matrix in this model, a concise analytical expressing is proposed, and some discussion about how to obtain the assignment matrix through simulation approach is made. In addition, the stochastic assignment matrix model proposed by K.Ashok is also discussed briefly in order to subsume all the newest methods to calculate the assignment matrix as we can. The assignment matrix model proposed in this thesis is more applicable for freeway's dynamic OD matrix estimation.