Converting Static to Dynamic Assignment Models: Preliminary Findings
VUREN T Van, CARMICHAEL S, Hague Consulting Group, HOFMAN F and TAALE H, Ministry of Transport, The Netherlands
Congestion keeps increasing, and will most likely keep doing so for the foreseeable future. Together with a change in the public acceptance of the impacts of traffic congestion on many aspects of life this has led to more prominent congestion-related tran
Congestion keeps increasing, and will most likely keep doing so for the foreseeable future. Together with a change in the public acceptance of the impacts of traffic congestion on many aspects of life this has led to more prominent congestion-related transport policy objectives. The time dimension has entered the thinking framework, not only through the reaiisation that time-of-travel choice has been a neglected aspect of individuals' overall transport decision-making, but also through a desire to make better use of existing transport facilities over the whole day. Appropriate modelling teclmiques are required, and there is a growing aspiration to use dynamic rather than static network models for forecasting and assessment. (In addition to these within-day dynanfics, there is a growing interest in day-to-day dynamics, reflecting daily variations in demand mad supply conditions, learning processes, and longer term transport dynamics, acknowledging the different time scales of changes in e.g. route, mode, destination and location choice; these are outside the scope of this paper).
Dynamic assignment packages have been around for two decades, and have been used sporadically for practical studies. Increasing computer power and continued developments in software have.resulted in easier access to these packages, whilst the wish to account for within-day dynamics has increased interest further. Parallel to this practical rising profile for dynamic network modelling, academic research has continued to improve our understanding of the defmitiun of dynamic equilibrium.
In practical reality there are many existing static network models in operation, which represent a high level of investment (in terms of data and staff expertise). It would be desirable to use this investment by converting static models to equivalent dynamic representations, aiming to maintain the main inputs and to stay close to the level of calibration achieved with the original, static model (at least at the time-aggregate level). Hence, conversion procedures are required. In addition, extra data inputs are required for the operation of dynamic network models. In particular, the fixed, hourly departure rate per OD pair, which suffices for static assignment models, must be broken down into a departure profile, so that the departure time process is properly reflected and the build- up and dissipation of queues can be modelled in more detail.
This paper reports on a projects that required conversion of static to dynamic representations, encompassing the QBLOK static assignment model (Bakker et al, 1994) and two dynamic models, CONTRAM (Taylor, 1990) and 3DAS (De Romph et al, 1992). The aim of the paper is twofold: to discuss the conceptual and practical issues involved in the transfer from an existing static assignment models to equivalent dynamic representations, mad to serve as a resource paper for those embarking on such an exercise.
This paper does not allow sufficient space for a detailed discussion of each of the assignment models involved. The references should be consulted for more comprehensive model descriptions.
Association for European Transport