Locating changeable message signs for advanced traffic information and management systems Liping Fu, Jeffrey Henderson, and Shuo Li Abstract: This paper presents an optimization model for locating changeable message signs (CMSs) on an integrated freeway-arterial network. Compared with existing models, the proposed model represents a well-balanced compromise between computational efficiency required to solve problems of realistic size, and model realism to ensure the quality of solutions. The model has three unique features: (1) it recognizes that locating CMSs is a planning problem that must take into account both current and future needs and benefits, (2) it evaluates benefits of CMSs over multiple time periods with different traffic distributions, and (3) it explicitly considers inherent variations in incident characteristics across links and over time. A sensitivity analysis is performed to examine the potential impacts on optimal CMSs locations resulting from uncertainties in various input parameters, such as traffic demand, incident attributes, and driver behaviour. Lastly, the proposed model is applied to the Highway 401 express-collector freeway system in Toronto for relocating the existing CMSs. Key words: changeable message signs (CMSs), location optimization, traffic assignment, queuing theory. Résumé : Cet article présente un modèle d’optimisation de l’emplacement des panneaux à messages variables (« CMSs ») sur un réseau général autoroutier. Par rapports aux modèles existants, le modèle proposé représente un compromis équi- libré entre l’efficacité computationnelle requise pour résoudre des problèmes d’une dimension réaliste et le réalisme du modèle qui permet d’assurer la qualité des solutions. Le modèle présente trois caractéristiques uniques : (1) il reconnaît que l’emplacement des « CMSs » est une question de planification qui doit tenir compte des besoins et des avantages actuels et futurs, (2) il évalue les avantages des « CMSs » sur plusieurs périodes et pour différentes répartitions de trafic, et (3) il tient compte des variations inhérentes aux caractéristiques des incidents quant aux liaisons routières et dans le temps. Une analyse de sensibilité est effectuée pour étudier les incidences potentielles des incertitudes de divers paramètres d’entrée tels que la demande de trafic, les attributs des incidents et le comportement des conducteurs sur l’emplacement optimal des « CMSs ». Finalement, le modèle proposé est mis en œuvre sur le système d’artères collectrices de l’autoroute 401 à Toronto pour changer l’emplacement des « CMSs » existants. Mots-clés : panneaux à messages variables (« CMSs »), optimization de l’emplacement, affectation de trafic, théorie des files d’attente. [Traduit par la Rédaction] Fu et al. 663 Introduction Changeable message signs (CMSs), also known as vari- able message signs (VMS), are becoming popular as one of the primary means for transportation agencies to disseminate travel and traffic information to motorists. Under the umbrella of intelligent transportation systems (ITS), CMSs constitute a key element in dynamic traffic management and informa- tion provision functions. The CMSs are commonly used to inform motorists of varying traffic, roadway, and environ- mental conditions and provide information on the location and severity of incidents and the expected delay. They can also be used to advise motorists of alternate routes in the event of an incident, construction, or a roadway closure. The effectiveness of CMSs, however, depends on how many CMSs are installed and where the CMSs are located in the network. Theoretically, the benefits from CMSs can be maximized, if the whole network of roads is instrumented with CMSs. This is, however, practically impossible because of the high costs of installing CMSs. Furthermore, past studies have suggested that excessive use of CMSs could lead to diminishing returns in benefits, and even worse, could be counter-effective because of the behavioural response of drivers to real-time information (Wardman et al. 1997). The planning of locations of CMSs is challenged by a variety of issues, such as how to model the response of drivers to changeable message sign (CMS) messages, how to model random incidents that vary by time and space, and how to model the impacts of incidents on traffic. The state-of-art practice has mostly relied on the experience and judgement Can. J. Civ. Eng. 34: 651–663 (2007) doi:10.1139/L06-151 © 2007 NRC Canada 651 Received 9 December 2005. Revision accepted 3 October 2006. Published on the NRC Research Press Web site at cjce.nrc.ca on 26 June 2007. L. Fu 1 and J. Henderson. Department of Civil Engineering, University of Waterloo, Waterloo, ON N2L 3G1, Canada. S. Li. College of Civil Engineering, Hunan University, Changsha, Hunan, People’s Republic of China. Written discussion of this article is welcomed and will be received by the Editor until 30 September 2007. 1 Corresponding author (e-mail: lfu@uwaterloo.ca).