117 Transportation Research Record: Journal of the Transportation Research Board, No. 1980, Transportation Research Board of the National Academies, Washington, D.C., 2006, pp. 117–125. Schemes are addressed for providing predictive travel time information in consistent anticipatory route guidance systems. A user-equilibrium time- dependent traffic assignment algorithm with the ability for multiple–user class network loading is used to generate consistent anticipatory route guidance information when a fraction of the population is equipped to receive and act on that information. Both a theoretical analysis and a simulation-based approach are presented to evaluate the provision of anticipatory route guidance information in terms of experienced path travel times and their variations. Results indicate that the proposed anticipatory route guidance strategy is accurate and reliable and can improve real-time traffic network performance. Advanced traveler information systems (ATIS) aim to relieve recur- rent and nonrecurrent congestion by providing trip makers with descriptive information (e.g., network conditions and available travel alternatives to help them make better travel decisions) or normative information (e.g., instructions to guide them effectively through the network to achieve some systemwide objective). Behavioral studies suggest that traveler acceptance of route guidance information depends considerably on whether such information is experienced to be valid (or consistent) and reliable (1–4). Although the effective- ness of traffic information has been extensively addressed in the lit- erature in terms of network performance improvement, the equally important issues of consistency and reliability have received only limited attention. Traffic information can reflect prevailing network conditions or predictive (anticipatory) network states. Prevailing information depicts current network (so-called instantaneous) conditions; it might be a summary description of traffic conditions at a particular time or path recommendations based on these prevailing conditions. A potential limitation is that network conditions may change signifi- cantly and hence invalidate path choice decisions made according to prevailing information. It has also long been known in the trans- portation science community (5–9) that route guidance based on prevailing or historical trip times could be counterproductive [i.e., worsen traffic conditions, partly because of the phenomenon that Ben-Akiva et al. call overreaction (9)]. Anticipatory information is derived from forecasts of network states. When predictions are accu- rate, anticipatory information is generally expected to be more effec- tive than prevailing information because it accounts for the rapid changes of traffic conditions spatially and temporally and is based on the traffic situation predicted to prevail at the time the trip maker reaches a particular location. However, the consistency between pre- dicted information and user reaction to the information remains an important research question that has been addressed only under somewhat restrictive simplified behavioral response assumptions. Recent results of simulation-based studies of an actual network by Mahmassani et al. confirm the potential improvement in effec- tiveness (in terms of overall network performance) that might be attained by providing anticipatory travel time information relative to prevailing travel time information (10). However, the reliability of the predictive information was not addressed in that study, and the question of consistency was addressed only to a limited extent. The present paper presents a more comprehensive evaluation of anticipatory travel times generated by a real-time traffic estimation and prediction system (TrEPS), with particular emphasis on the reli- ability of the predicted path travel times from an individual traveler’s perspective. The generation of predictive travel times is considered as the consistent anticipatory route guidance (CARG) problem, which is formulated mathematically as a fixed-point problem and solved by the user-equilibrium time-dependent traffic assignment (UETDTA) algorithm proposed by Mahmassani and Peeta (7 ). Both analytical and simulation-based approaches are used to evaluate the generated anticipatory route guidance information in the form of predictive point-to-point travel times. Although the main focus is CARG reliability, other important properties (e.g., consistency and effectiveness) also are examined. In the analytical approach, the traffic network is modeled as a queueing network in which each signalized intersection is modeled as a server and each approaching vehicle as a customer arrival. The simulation-based approach uses a closed-loop rolling horizon frame- work in which the simulation assignment model, DYNASMART (11), is embedded to estimate current traffic conditions and to pre- dict short-term future traffic conditions. In both approaches, the reliability of predictive travel times is measured in terms of the max- imum difference in the experienced path travel times over all the used paths connecting each origin–destination (O-D) pair for each departure time interval. Furthermore, the consistency of predictive travel times is examined by comparing the predictive travel times with actual user travel times. For completeness, relative network perfor- mance (in terms of average trip time) is compared under prevailing and predictive information provision scenarios to demonstrate the effectiveness of predictive travel times. The paper is structured as follows. First, a fixed-point formula- tion for generating CARG information and the corresponding solu- tion algorithm are presented. Next, the analytical and simulation approaches to evaluate the anticipatory route guidance information How Reliable Is This Route? Predictive Travel Time and Reliability for Anticipatory Traveler Information Systems Jing Dong, Hani S. Mahmassani, and Chung-Cheng Lu Maryland Transportation Initiative, Department of Civil and Environmental Engineering, University of Maryland, Martin Hall, College Park, MD 20742.