Urban Traffic Networks

This paper addresses the problem faced by a central controller seeking to optimize overall network performance through the provision of real-time routing information to suitably equipped motorists. Conceptual and mathematical formulations are presented for various scenarios that arise based on the amount of information available to the controller. Principal elements of a dynamic assignment formulation for electronic route guidance systems are discussed, and the associated difficulties for solution methodologies are illustrated. The ideal case of known time-dependent origin-destination flows over the whole planning horizon is formulated as a dynamic system-optimal assignment problem. Extensions and variants of the basic formulation are discussed for incomplete information availability to the central controller.

Higher levels of market penetration of dynamic route guidance systems, such as the EURO-SCOUT system from Siemens, require that some form of adjustment is made for the consequences of routeing recommendations. Ideally a procedure is required that generates route recommendations such that no guided driver could have taken a faster route. This is a form of Wardrop equilibrium, and necessitates a dynamic equilibrium assignment algorithm in order to obtain such a set of route recommendations. The determination of a dynamic equilibrium assignment requires the derivation of travel time functions. In this paper it is proposed that link travel time measurements from the dynamic route guidance system be combined with link flow measurements from a traffic signal control system, and that the coefficients of the link travel time functions be estimated recursively. A modified Frank-Wolfe algorithm. for solving the dynamic equilibrium assignment problem is proposed. This requires the iterative use of a shortest path algorithm, such as that currently implemented in the EURO-SCOUT system. The properties of the algorithm are explored using a small numerical example.

The acceptance of route guidance advice by motorists is expected to occur only whenever such advice is experienced to be valid and reliable. Two factors may cause route guidance systems to provide motorists with unreliable advice: (1) the traffic information constituting the basis for guidance advice is inaccurate; and, (2) the impact of a large fraction of motorists responding to the guidance, and the subsequent overreaction that occurs, is ignored whenever the guidance advice is being set.

This paper specifies the overall structure of an on-line Dynamic Traffic Model System for an inter-urban motorway network. The model predicts traffic conditions on a motorway network in real time, and can be used to provide information on (expected) congestion to a motorway traffic control centre. The model receives traffic data such as traffic flow, average speed and average detector occupancy in real-time from a motorway monitoring system. Additionally, a database of historic traffic information will be used.

In urban traffic planning and management the determination of network configuration (topology and link capacities), signal setting parameters (cycle time, green times and offsets) and link flow pattern is generally required. To this aim a set of models, belonging to the family of network design models, has to be defined. In these models the path choice user behaviour has to be taken into account, in order to define the link flow pattern. The signal setting contributes to the definition of the user costs (travel and delay costs) and assumes a relevant role in the path choice. For this reason the signal setting parameters (data and variables) should be explicitly included in the model formulation.

The industrial development of the last years and its interaction with the environment is one of the most important problems of the next future. While the pollution due to industrial production is actually strictly regulated by laws that force companies to observe precise constraints, pollution due to vehicle emissions, because of pollutants sources general diffusion, is a problem for whose solution technological and political measures conditioning collective customs and manners are needed. The adoption of these measures, especially for emergency situations, is performed on the base of data provided by special detectors capable of measuring concentrations of some substances. This methodology can cause two great problems: (i) because of required equipment high costs, it is not always possible, on the ground of fixed detectors location in a certain area, to obtain reliable information on the distribution of the pollutant substances on the whole area; (ii) mistakes strongly depending on receptor localization can arise, for instance when detectors are located in proximity of pollutant sources. In order to solve these problems, general models for the evaluation of pollutants levels which take into account the different complex factors (i.e. traffic data, meterological conditions, topological characteristics of the observed area) influencing the phenomena of the emission and of the diffusion of the pollutants are needed.

The purpose of this paper is to illustrate the effects on estimated traffic flow of different traffic models in solving the dynamic traffic assignment problem. We will compare two distinct traffic models, the kinematic wave model and the flow model. We will show that the model’s resulting estimates differ substantially. The kinematic wave model is regarded as the more realistic of the two but it is difficult to work with, requiring large computational effort, while the flow model is less realistic but is computationally more tractable. The flow model has been adopted widely in the development of existing dynamic traffic assignment techniques (Wie, Friesz and Tobin, 1990; Ho, 1990, Vythoulkas, 1990a; 1990b; Carey, 1986; 1987; 1988). According to this model, the outflow from each link is a function of the mean density on the whole link.

This paper proposes a time dependent assignment model based on a rigorous formulation of dynamic equilibrium that aims to equilibrate real travel times. The paper also describes the algorithms devised to solve this problem and their tests on simplified networks

We present a vehicle-by-vehicle deterministic queueing traffic assignment model for minimising total travel delay in a congested road network. The model routes drivers, one at a time, according to the marginal delay of each link. This is regarded as a combination of the delay caused by each driver to others travelling in the whole network, and the travel delay experienced by the driver himself. This yields an approximate system optimal routeing pattern and the corresponding road prices needed to cause the user equilibrium traffic pattern, which would arise from each driver minimising only his own travel delays, to be the approximate system optimal one. A key characteristic of the model presented is that it is applicable to (multi-commodity) networks having many origin-destination pairs and many bottlenecks. The model has evolved as a result of the initial study on this problem in Ghali and Smith [1], and basically extends the model in that paper. Computational results comparing network performance of applying the model of this paper against that of the previous paper as well as against network performance due to the natural user equilibrium networks are provided.

In order to assess the performances of information systems to road users and of demand management measures, it is necessary to simulate dynamics of traffic flows on the transport network and users’ behaviour. This can be accomplished by taking into account several aspects such as users’ information level, travel demand distribution over time and the expected smoothing effect of adopted measures and information systems. A computational procedure to perform this assessment for various traffic scenarios and hypotheses about users’ behaviour is presented in this paper. Moreover, the indicators needed to quantify information systems and demand management effectiveness are identified together with parameters defining traffic scenarios. The results of an application of the procedure as a deterministic simulation tool are also presented and discussed.

In this paper an algorithm to calculate extremals for the dynamic traffic assignment problem (DTAP) using a deterministic optimal control problem formulation is shown. This algorithm is based on a general method to integrate extremals for optimal control problems with inequality constraints in control variables. The DTAP is formulated and discretized and the consideration of using the method for extremals integration is made, showing how its steps would result for the DTAP, specially its inner loop, that becomes a network flow problem very similar to the stationary traffic assignment with separable costs objective function. Also, for the algorithm, it is remarked the need of techniques to update solutions in traffic assignment problems when, on the same network and the same cost functions, the demand varies slightly from one problem to the next. Finally aspects of the algorithm still under study are outlined. These aspects must be specialized for the DTAP in order to make possible the writing of competitive code specially suited to the treatment of medium-large urban networks

Congestion costs and optimal tolls have long been computed from static (single period) traffic assignment models. Only recently have efforts been made to derive congestion prices from dynamic traffic assignment models (i.e., models having time-varying traffic flows). We consider this problem here, and find that in the dynamic context, congestion costs imposed by users consist of a capacity cost plus a component associated with the first-in-first-out (FIFO) property of road traffic.

This paper presents an equilibrium assignment methodology that can obtain delay values from a simulation program. The proposed assignment procedure is compared with the Frank and Wolfe algorithm with fully specified volume-delay functions and with the method of successive averages (MSA) suggested by Sheffi to be used in stochastic assignment problems. Some results of the application of the proposed algorithm on a real network are given, and a proof of convergence for the simple case of a two-link network is provided.