Moment-based travel time reliability assessment with Lasserre’s relaxation

Abstract

ABSTRACTThis paper proposes the generalized moment problem (GMP) to assess total system travel time (TSTT) reliability on a traffic network. As a distribution-free method, the proposed approach only needs moment information of uncertainty to make the evaluation. In the GMP, upper bound of the probability that TSTT exceeds some predefined threshold value (which is called unreliability function) is the objective, and moment information of uncertainty is formulated as constraints. Based on the first moments and support information of total link travel time of each link, we use Lasserre’s relaxation to reformulate the GMP as a semi-definite programming, which can be solved efficiently. We test the proposed method on traffic networks of different sizes. Numerical results show that the upper bounds obtained with our method are tightest in most cases and show different tail information. The results may also bring some useful insight to applications of the moment-based method in practice.