Non-Recurrent Congestion Essentially, non-recurrent congestion is the cumulative effect of multiple actions in response to a new environment. An individual may be highly familiar with their usual route from work to home, say, when faced with an incident they must make a different choice. They must individually identify the best way from their current location to home, considering the new conditions. These route alterations, away from the norm, cause space conflict and congestion to form in unforeseen areas of the road network. This leads to a change in network performance, as is demonstrated conceptually in the diagrams below: Current Modelling Approaches As indicated by the nature of this type of congestion, this form of congestion requires careful modelling of driver responses to the new situation. This research paper, therefore, initially sought to examine the current range of traffic models and their suitability for this task. Traffic simulation is best distinguished according to the scale at which the simulation is focussed. At one end, macroscopic simulation models the collective movement of vehicles, whereas microscopic simulation represents traffic from the perspective of the individual vehicle, simulating the psycho-physical attributes of moving vehicles on a highly detailed basis. Mesoscopic simulation incorporates aspects from both approaches. The following table summarises these approaches: Despite the range of such software available, in many respects there remains a concentration upon a core of underlying theories and techniques. The assumptions made here are of particular importance for the purposes of this research project. . Irrespective of the scale chosen at which to simulate traffic, existing simulation tools utilise a core of modelling practices. Of most importance, to the research case being addressed here, is the handling of driver route choice. As described earlier, the redistribution of traffic in response to irregular environments is a cumulative effect of driver choice. The choice aspect within existing models, however, has been largely simplified. The majority of models, independent of the scale of simulation being adopted, treat this factor from a macroscopic perspective. Traffic flows are ‘assigned’ according to principles of equilibrium described by Wardrop (1952). While these models are strong representations of everyday normal behaviour, where drivers know the best route between origin and destination, they make unreasonable assumptions regarding route knowledge and behaviour in irregular environments. The nature of the environment that is being described here means there is a greater necessity for accurate representation of the variety of behaviour on the road network. Such actions can not be easily generalised, and so approaches that seek to capture the full complexity of the environment should be examined. Driver Cognition and Decision-Making The final area of research begins to examine the way in which drivers make route choices in irregular environments. In replanning their journey in response to an incident, a driver will utilise their existing understanding of their environment. It has been shown that people do not necessarily incorporate a Euclidean spatial understanding when making these choices. Instead space is understood in terms representations similar to the ideas behind cognitive maps (like that shown in Figure 2). This understanding reinforced by a hierarchies of objects, particularly landmarks, where certain items are more important than others. A number of authors (notably Hirtle & Jonides 1985) have demonstrates the importance of landmarks in navigation and wayfinding, rather than the shape of the road on which they travel. Simulation of Non-Recurrent Congestion on Urban Road Networks Supervisors: Dr Tao Cheng Alasdair Turner Andy Emmonds Ed Manley MRes Urban Sustainability and Resilience Department of Civil, Environmental and Geomatic Engineering, Gower Street, London WC1E 6BT Sponsors: Road congestion is a familiar urban phenomenon – blocked roads, slow journeys and frustration. It is damaging to the economy, health and the environment. The Department for Transport predicts that excess delays in urban areas costs the economy £10.9bn a year, while contributing to long-term respiratory problems of city dwellers (Department for Transport 2009). The problem is spreading too, as countries develop economically and more people buy cars, the roads become more congested. People, politicians, planners and environmental workers are all seeking an answer to the congestion conundrum. Yet the majority of congestion (around 55% according to the Federal Highways Administration (FHWA 2004) and other sources), is caused by irregular events in the roadway. The reduction in capacity caused by an unexpected incident upsets the day-to-day equilibrium that characterises urban traffic flow. The systemic response to such incidents is known to be ad hoc and difficult to predict. This poster presents an overview of a research thesis conducted into the necessity and feasibility of a simulation model for non-recurrent congestion. The literature review examined in depth the limitations of current techniques in examining this phenomena, and presented a number of alternative strategies that may shed new light on the problem. In essence, the thesis advocates the use of individual-based modelling approaches in seeking to understanding global phenomena. As will be demonstrated, the nature of non-recurrent congestion requires a more careful understanding of the behaviour of individuals, and hence guides the approach described. Scale of Simulation Macroscopic Mesoscopic Microscopic Simulated unit Traffic flows Traffic flows with vehicular route and node interaction Individual vehicle behaviour Fundamental laws drawn from Physics, Mathematics Physics, Mathematics Cognitive studies, Physics, Artificial Intelligence Input volume of data Low Medium High Computational need Low Medium High Scale suitability Small scale analysis Very poor Medium/Poor Very strong Medium scale analysis Medium Strong Strong but computationally intensive Large/City scale analysis Very strong Medium Too computationally intensive Commercial exposure Key commercial applications TRANSYT, FREQ12, Synchro DYNASMART, DYNAMIT, CONTRAM AIMSUN, VISSIM, CORSIM, Paramics Modelling Complexity Complexity theory represents a swathe of approaches towards the understanding and modelling of complex phenomena. They seek to replicate the way in which situations arise through the interaction of individual entities. Such interactions can, depending on the scale at which the system is viewed, produce so-called emergent properties. These are properties of a system that exist solely through the interaction of individuals, they can not be attributed to the properties of any individual. It has previously been shown how congestion is an emergent properties of the interactions between individual vehicles on the road network (Manley and Cheng 2010). There exist a number of modelling approaches that offer promise where the simulation of such properties are concerned. Of greatest note, with respect to this project, is that of Agent-based Simulation. In a similar respect to microscopic traffic modelling, agent-based simulation models an environment from the individual unit upwards. However, vitally they differ in terms of the treatment of behaviour modelling. Agent-based modelling allows a far greater degree of autonomy within the individual, and an individual-based intelligence that enables them to cognate the world around them. This allows one to build a picture of global behaviour from the individual behaviours of a collective. Within regard to the situation being discussed, this intuitively has significant implications for the representation of non-recurrent congestion. Building this model of global behaviour requires careful selection of the type of individual action that may contribute to it. With respect to non-recurrent congestion, as has been discussed, of great importance are the new routes that people pick in response to irregular events. By modelling the collective behaviour, of this type, in these environments, one may be able to build a picture of the movement of congestion around the network. In addition to models of spatial cognition are those that replicate the way by which people choose their routes. Such heuristics replicate the process by which people utilise their internal maps and hierarchies in navigating their way around a city. These types of behaviours are, of course, important when responding to irregular incidents. Although there remains further research to be done in this respect, it is hoped that these modelling approaches represent a route forward in the modelling of global phenomena from individual choices. References FHWA, 2005. Traffic Congestion and Reliability: Trends and Advanced Strategies for Congestion Mitigation. Available at: http://ops.fhwa.dot.gov/congestion_report. Hirtle, S. & Jonides, J., 1985. Evidence of hierarchies in cognitive maps. Memory & cognition, 13(3), 208-217. Lynch, K. 1960. The Image of the City. MIT Press. Manley, E. & Cheng, T., 2010. Understanding Road Congestion as an Emergent Property of Traffic Networks. Proceedings of International Multiconference on Complexity, Informatics and Cybernetics. 1. 109-114. Wardrop, J. C., 1952. Some Theoretical Aspects of Road Traffic Research, Proceedings of Institution of Civil Engineers, Part 2, 9, 325–378. Figure 2: Sketch Cognitive Map of Boston, USA (Lynch 1960) Table 1: Comparison table of traffic simulation methods Conclusions and Future Directions The research undertaken here represents the first stages towards a simulation of non-recurrent congestion. It has been demonstrated that many existing traffic simulation methods do not carry enough behavioural complexity to enable the effective modelling of such global behaviour. However, there remains great potential for those tools arising from the field of complexity to fill this void. Initial research has suggested that the modelling of driver cognition and route decision- making behaviour holds some potential, with a large body of research already existing in this area. This will henceforth be further pursued, with prototype modelling carried out to demonstrate the validity of these ideas. The overall ambition of this work is to produce a more realistic representation of global behaviour on the roads in response to irregular environments. Blockage Prior to Event Post Event Direction of Travel Congested Heavy Medium Light Traffic Volume A Road B Road Residential Road Size Figure 1: Changes to whole network performance in response to irregular event