Modelling and Control of Multiple User-class Freeway Traffic Flow by Using Dynamic Lane Assignment
HOOGENDOORN S and BOVY P, Delft University of Technology, The Netherlands
In modem traffic management, distinction in the allocation of infrastructure to different user-classes is proposed and used as a measure to make more efficient use of(existing) infrastructure facilities. The principle underlying this policy is to grant pr
In modem traffic management, distinction in the allocation of infrastructure to different user-classes is proposed and used as a measure to make more efficient use of(existing) infrastructure facilities. The principle underlying this policy is to grant priority to socially or economically important user-classes (e.g. freight transport services, car- pnolers, law-enforcement authorities, or traffic travelling in a specific direction) with respect to other user-classes. In this respect, research has shown that while freight- transport and business-related traffic only represent 6%, and 24% of the total traffic demand respectively, they account for 22%, and 47% of cost inflicted by travel time losses due to congestion (see [Stern et al.,1995]). This example encourages priority policies on economical grounds.
In general the objectives of priority-policies should aim to endow genuine and efficient benefits, preferably comprising the whole range of transportation operations of the user-class. Within these more comprehensive (network-wide) measures, user-class- dedicated lanes can be seen as an individual measure, part of an integral transportation application.
The numerous examples of fixed, that is not reactive to a changing traffic demand and composition, allocation polices found in practice involve among others car-pool, bus-, truck-, tidal-flow and shoulder-lanes. However, these fixed configurations cannot react to a changing traffic demand and composition. It is expected that a more efficient utilisation of infrastructure may be obtained by employing dynamic lane assignment (DLA) policies.
The impact of DLA policies can be evaluated using objective fimctions. For general traffic management applications, these objective functions are functions of different measurable indicators (e.g. effective capacity, travel time losses due to congestion), reflecting the (quantitative) effect of the application on different user-classes affected by the application. In general, objective functions are used to assess the dynamic lane allocation policies.
However, objective functions can also be utilised as an input for automated control schemes, generating optimal control policies (with respect to the objective function).
In this paper, Model Based Predictive Control (MBPC) is called upon for the automated generation of optimal DLA policies. MBPC is an advanced controlling technique, featuring advantages ofbothfeedforward and feedback control As the term model based suggests, a model capturing (at least) the essential characteristics of the underlying dynamic process should be available. Due to the envisaged functioning of the model, a basic requirement imposed on the model is the distinction between the various user-classes of interest, and the lanes of the road. Moreover, for the model to describe real-life traffic operations realistically, knowledge about the exogenous inputs (disturbances) is required (e.g. dynamic traffic demand and composition, other disturbances affecting parameters in the model, like weather conditions, choice behaviour of drivers, and capacity at the exit of the freeway section). Possibly, additional constraints concerning the resulting dynamtc control configurations and characteristics need to be considered.
Association for European Transport