Contents I. Introduction and Background
This paper is concerned with the problem of validating a macroscopic traffic flow model along the lines given in [1] and [2]. In [3] a flexible optimisation formulation was introduced for the calibration problem of the macroscopic simulator Modéle d’Écoulement de Trafic sur Autoroute NETworks (METANET) [4]. It features the arbitrary selection of fundamental diagram (FD) application range for homogeneous road stretches, which themselves are split into segments, but also penalised the variance of their parameters. The rationale behind this penalisation was for obtaining solutions favouring similar FDs over a road network. This kind of similarity was employed as guidance when validating the large scale model of the Amsterdam motorway network, [5], [6], but the implementation required resorting to intuition and engineering expert opinion. Here, by considering in the problem formulation the FDs as extensive quantities spanning over a number of motorway links, the use of experts’ knowledge is replaced by the optimisation process. This further supports the automation of decision making regarding the model validation procedure and the development of systems that are able to self-calibrate and maintain themselves, as suggested in [7] and [8] for macroscopic and in [9] for microscopic models. Within this framework, the work reported here supports the feasibility of the automatic build-up of libraries of models based on data from different dates, assuming a proper interface with the traffic data collection and archiving system. The relevance, transferability and applicability, though, of an identified set of model parameters at particular situations and time of day, would be something considered at a higher level.
