The dynamic estimation of Origin-Destination (O-D) movement in traffic networks potentially forms an important component of traffic guidance and control systems, such as ramp metering and variable message sign systems, as well as providing an important source of demand information of use for longer term transport planning. This paper reviews a number of approaches to the problem that are currently under consideration by the ODIN consortium (DRIVE project V 1047). The first approach covered is discrete time recursive (and non-recursive) estimation for applications where measurements of traffic flows are continuously made on the entrances and exits to the network and where movements between entrances and exits are to be estimated. In this situation, traffic assignment information is not required. Extensions to these techniques to allow for journey time distributions are described. Discrete time estimation requires the aggregation of counts over time intervals, leading to the loss of information. An alternative fully disaggregate approach which also allows for journey time distributions is described. For applications where traffic flow measurements are available for links within the network, traffic assignment information is required (even where there is no choice of route). Estimation techniques which allow for time-dependent assignment are described and an efficient "rolling horizon" sequential estimation algorithm is presented.
Abstract