A conceptual framework for specifying and developing pedestrian models

Pedestrian modelling is frequently used for making decisions regarding the planning, design, and management of pedestrian areas. For example, the designer of a new shopping mall would be interested in what locations people are likely to be attracted to, or the operators of a large-scale event might like to know where congested areas are likely to occur so they can develop management plans. The outputs of thesemodels can include flows on certain routes, entry and exit counts, and level-of-service graphs. Computational modelling of pedestrians is sometimes difficult due to the complex and random nature of pedestrian movement. Pedestrians often make unconscious decisions that are difficult to explain or measure. They move at a much smaller scale and in a less constrained manner than other vehicles, meaning techniques developed for modelling other modes of transport cannot be translated to pedestrians easily. Pedestrians interact with many different kinds of environments. Enclosed spaces can consist of rooms connected by corridors leading to and from exits (eg. office buildings, shopping centres) or can be more open-plan (eg. sports arenas, train stations). Mixed mode environments consist of areas, possibly shared with cars or public transport, which connect the pedestrian to building entrances and other streets. Open areas consist of open areas and/or designated pathways where pedestrians wander and sightsee. Some environments are a hybrid of the above, for example sports precincts or universities, and generally include pedestrian areas or low vehicle traffic areas containing several attractions. There are also several different types of behaviour exhibited by pedestrians in these environments. Some pedestrians know where they are going, how to get there, and has a very low probability of being distracted on the way (purposeful and familiar). Some know where they want to go, but is not sure how to get there and as a result may get distracted or lost on the way (purposeful and unfamiliar). Sometimes pedestrians have no purpose and are just wandering. On rare occasions, pedestrians may be in panic mode and will behave differently to normal. Two other behaviours are forced waiting, where pedestrians need to wait for an environmental action before continuing (eg. waiting in a queue or waiting for traffic lights to change), and also the incorporation of temporal constraints into their planning (eg. being on time for a train). A range of techniques have been used for pedestrian modelling. Mathematical approaches typically use differential equations to model the speed and location of pedestrians. Alternatively, cellular automata models employing simple update rules have also proven to be useful. However, the discrete nature of automata based-models reduce their functionality for some applications. Agent-based modelling and simulation has also been explored with some success. Finally, simulations based on aggregate traffic modelling techniques have been used in industry. Many stakeholders are involved in the development of a pedestrian model, including the client, the practitioner, and the developer. The client requires results that they can incorporate into their ecisionmaking process and communicate to others. The practitioner requires a model that is easily adapted and models the environment closely. The developer must create a model that meets their requirements. The main factor in selecting an approach is the location and intended use of the area to be modelled. Other factors include the behaviours required and the scale of the model. We discuss the strengths and weaknesses of each approach for each environment and proceed towards a framework for selecting an approach for the intended application. This framework will be of use to clients, practitioners, and developers. It will play a strong role in the usefulness and reliability of pedestrian modelling in the decision-making process for planning and design of pedestrian-frequented areas.