Transportation Research Record: Journal of the Transportation Research Board

The existence of equilibrium solutions in single-bottleneck models with homogeneous travelers having the same preferred arrival times is explored from both theoretical and experimental frameworks. First, an equal payoff solution obtained in past studies is proved not to be in dynamic user equilibrium (DUE). Necessary conditions for DUE are presented, and the nonexistence of DUE in a single-bottleneck model with homogeneous commuters is proved. Second, a multiplayer online experimental network game that emulates the single-bottleneck model is discussed. The game is played online by real players and is repeated 21 times, each under no-information and online-information settings. The experimental setup and the results are described with insights from graphical plots and analysis. Specific emphasis is placed on the system's dynamic evolution in each game and the pattern of a user's decision making. It is anticipated that these results will provide important insight into the underlying behavioral decision making of travelers and will thereby facilitate the development of better dynamic traffic assignment models.

A new analytical multiclass dynamic network loading (DNL) model is proposed as part of a simulation-based dynamic traffic assignment (DTA) model. In contrast to many other proposed DNL models, this model explicitly deals with queuing and spillback without relying on link travel time functions as input. Different vehicle types (e.g., cars and trucks) are considered with different speeds. Use of link travel times is likely to under-or overestimate the true travel times in a dynamic model if queues are considered. The proposed DNL model consists of a link model and a node model. The link model computes queue inflows and potential outflows, whereas the node model determines the actual outflows depending on the node structure. In the end, the link travel times are computed backward in time for each vehicle type separately. The model has been implemented in the Indy DTA software, and an application shows that the approach is viable in real-life networks.

The method of successive averages remains by far the most widely used solution heuristic in simulation-based dynamic traffic assignment. Its simplicity and the nonrequirement of derivative information for the flow-cost mapping function are the main reasons for its widespread use, especially in the realm of dynamic traffic assignment (DTA). However, its convergence properties in real-life networks have been inconclusive, especially because (a) simulation-based models typically are not well behaved mathematically, and therefore their solution properties are not guaranteed, and (b) predetermined step sizes do not exploit local information in searching for a solution and therefore tend to have sluggish performance properties. An effort was made to improve on the performance of the method of successive averages heuristic for user-equilibrium and system-optimal DTA problems on large congested networks through novel implementations that derive their efficiency from exploiting local information made available in the results of vehicle-based simulation models used to provide the mapping between a feasible path flow assignment and the experienced travel cost in a DTA solution framework. The results of extensive numerical tests on actual networks are reported, confirming the performance improvements attainable with the new approach.

The availability of advanced traveler information systems technologies has provided opportunities for travelers to obtain online travel information and thereby improve their travel experiences. The resulting change in user behavior because of such information needs to be accounted for in traffic assignment models. A methodology that accounts for user recourse in the traffic assignment problem is proposed. This methodology accounts for the heterogeneity in user information perception by using a sequential form of logit model. Hyperpaths instead of elementary paths are used in this approach. A key difficulty in modeling online recourse in network assignment problems is in obtaining efficient hyperpaths. A methodology that uses a sequential form of logit model that obviates explicit enumeration of hyperpaths is proposed. Computational results from a hypothetical transportation network are presented to demonstrate the application of the proposed assignment approach.

Properties of a doubly dynamic simulation assignment model were investigated. The model involves specifying a day-to-day route choice model as a discrete time stochastic process, combining a between-day driver learning and adjusting model with a continuous-time, within-day dynamic network loading. Such a simulation model may be regarded as the realization of a stochastic process, which under certain mild conditions admits a unique stationary probability distribution (i.e., an invariant probability distribution over time of network flows and travel times). Such a stationary state of the stochastic process is of interest to transport modelers, as one can then describe the stochastic process by its moments such as the means, variances, and covariances of the flow and travel time profiles. The results of a simulation experiment are reported in which the process of individual drivers' day-to-day route choices are based on the aggregate learning of the experienced within-day route costs by all drivers departing in the same period. Experimental results of the stationarity of the stochastic process are discussed, along with an analysis of the sensitivity of autocorrelations of the route flows to the route choice model parameters. The results illustrate the consistency of the link flow model with properties such as first in-first out, and a simple network is used to illustrate the properties.

Uncertainties are unavoidable in engineering applications. A new model is proposed for designing networks under uncertainty of future demands. The objective is to minimize the total travel time budget required to satisfy the total travel time reliability constraint while considering the route choice behavior of network users. The model adopts the value-at-risk risk measure instead of the utility function to model planner risk preferences. It allows the planners to specify their risk preferences by using a confidence level of alpha on the total travel time reliability. This alpha reliable network design model is formulated as a stochastic bilevel optimization problem. The upper-level subprogram is a variant of the chance-constrained model that minimizes the total travel time budget subject to a chance constraint with a user-specified confidence level, a budget constraint, and design variable constraints; the lower-level subprogram is a user-equilibrium problem under demand uncertainty. A simulation-based genetic algorithm procedure is developed to solve this complex network design problem (NDP). Two numerical examples are presented to illustrate the features of the proposed NDP model.

A robust optimization model is presented for the dynamic traffic assignment-based continuous network design problem, which accounts for a bilevel objective and long-term origin-destination demand uncertainty. The model also embeds Daganzo's cell transmission model. The objective minimizes the trade-off between expected total system travel time (TSTT) and expected risk. As such, the robust model provides the optimal solution that is least sensitive to the variation of travel demand, given the degree of robustness by transportation planners. The new robust model is compared with the existing network design models on a simple cell transmission test network. The robust model with greater degree of robustness yields less expected risk with the sacrifice of higher expected TSTT. The robust model yields the most robust solution, and no other model provides a satisfactory solution across the budget range. In addition, how a visualized graph may be used to elicit the preference information from transportation planners on the desired degree of robustness is illustrated.

The Dutch national road network has been developed over centuries. In the past, roads were constructed according to contemporary spatial and transportation planning philosophies. Because the existing road network is a result of a long process of successive developments, the question can be asked whether this network is the most appropriate currently, especially considering the socioeconomic structure of the Netherlands. To answer this question, an optimization algorithm has been developed for designing road networks for a given spatial structure of cities, villages, and industrial activities. With this algorithm, the Dutch road network is redesigned on the basis of minimization of the travel and infrastructure costs and by taking into account the socioeconomic structure of the Netherlands. A comparison between the existing network and the new design shows that the redesigned Dutch national road network has significantly lower total costs than the existing road network. It is found that the construction of roads with more lanes on somewhat different locations leads to a reduction of the total travel time and the total vehicles kilometers traveled.

In transportation engineering, the network design problem (NDP) aims at finding an optimal link improvement in a network for given demand. Although traffic demand is essentially uncertain, many previous studies assumed it to be deterministic. The demand stochasticity is incorporated and formulas are developed for system-optimal (SO) flow stochastic NDP. The SO assumption is justified by comparing results from SO deterministic NDP with those of user-equilibrium NDP. The difference in social cost between the two approaches is found to be less than 5%. Two two-stage stochastic programs with recourse formulations are proposed: one with penalty function and the other without. The main advantage of the first formulation is that a planner can exert better control on improvement by appropriately weighing reduction in the congestion versus improvement costs. The challenge, however, lies in selecting an appropriate penalty function. A nonlinear penalty function is found suitable for the test network studied. The second formulation does not require penalty function but results in a large number of scenarios. Nonanticipativity constraints are introduced in the second formulation to arrive at uniform improvement over all scenarios. Both formulations are solved on a test network. It is found that necessary improvements and the total costs with both models are more than those for average demand.

A study was done to develop a shortest path algorithm in which the realism of way finding had been improved by incorporating the concept of road cognition. The existing shortest path algorithms consider travel time and travel distance only. However, cognition of road configurations also can be considered for providing a realistically satisfied path. Surveys have shown that people prefer a high-cognition path rather than a low-cognition path. Therefore, the study developed a shortest path algorithm to provide a realistically plausible route that is preferred by many people and that considered the travel time, travel distance, and road cognition. The profiles of travel times, which are required to estimate arrival times up to destination, are predicted by using a Kalman filtering technique to reflect time-varying travel conditions. The travel distance is calculated by using the geographic information system (GIS) attributes, and road cognition is established by applying a space syntax technique to identify road configurations. Also, space data of the space syntax model are based on GIS data. By considering both the shortest path in terms of travel time and distance and the best cognition in terms of road configurations, multiple paths are determined.