This paper aims to describe and analyze the error of a model and to provide detection criteria and corrective treatments. The paper puts forward a systematic certification method which is valid for any model. The method is applied to a route choice model on a transport network, with a continuous distribution of user values of time. The discrepancy between the model and the studied system can have several causes: it may be an error in design, or in formulation, or in computation, or in estimation, or even an exogenous error which is propagated through the model. These different causes all require specific detection and treatments, it is therefore necessary for each of them to be clearly identified and defined. to be clearly identified and defined. The paper contains three sections. The first section distinguishes between the different causes of error and shows how they combine with each other. Next, numerical details concerning the exogenous error in a traffic assignment model are provided. This error could relate to origin-destination flows, travel time functions, or behavioural parameters such as the value of time. Lastly, a method for tracing the propagation of the exogenous error through a dual criteria assignment model is defined. In the context of a realistic interurban road traffic study, it is shown how the exogenous errors contribute to the uncertainty in the results of the model.
Abstract