Modelling Walking Behaviour
The main objective of my research is the implementation of a walker model that is accurate and able to simulate pedestrians walking in large walking facilities.
During the research we implemented, improved, extended, calibrated, validated and applied the Nomad model. Nomad is a microscopic pedestrian proposed by Prof. Serge Hoogendoorn.
The research objectives can be summarised in:
- Modelling pedestrian activities found in large walking facilities.
- Development of a simulation microscopic model for the application in large facilities.
- Developing implementation algorithms to improve computational efficiency.
- Development of a calibration and validation methodology.
- Validation of the walking behaviour in Nomad using the methodology.
- Face validation of the pedestrian activities by performing several case studies.
The movements that pedestrians perform are complex and pedestrian flows show a wide range of phenomena from individual to collective patterns. Pedestrians are capable to perform very high accelerations when initiating movement or when stopping. They can walk in several directions including backwards and sidestepping. Self-organised collective patters such as lanes that are formed in bidirectional flows or stripes in 90 degree crossing flows appear without coordination and are important for the traffic characteristics of pedestrian flows. Behaviours vary according to the composition of the population and their cultural backgrounds.
These complexities in pedestrian behaviours pose big challenges in the development and testing of pedestrian models created to predict these behaviours. Models that have a good predictive capacity and simultaneously provide a platform for exploration of human behaviours are more valuable to science. We take the view that this is also valid for the science of pedestrian modelling.
Given the impossibility to determine all the subjective factors affecting the walking behaviour the development of models is facilitated if they are coupled with a theory of pedestrian behaviour. In our research we developed and investigated the Nomad model that is derived from a normative pedestrian theory. The theory assumes that pedestrians are optimisers of the utility gained when performing activities and the utility lost when walking. The concept of utility can be viewed as an extension of the more popular concept of “effort” frequently referred in the principle of effort minimisation. Utility gain and loss can have objective and subjective origins. Objective utility originates from minimising the walking time or walking distance. Subjective utility incorporates preferences for certain activity areas, attractions to specific walking areas such as shopping windows or pleasant streets.
To develop and implement pedestrian models that are used to investigate and predict pedestrian behaviour we synthesise the requirement of pedestrian models in three scientific dimensions: the fidelity of the behaviours, the accuracy of the outcomes and the functionality available (which behaviours the model is capable to reproduce).
Fidelity: The large and complex range of pedestrian behaviours, reflected in their individual movements and collective patterns must be correctly represented by pedestrian models. Pedestrian models must be able to predict correctly self-organised patters and the expected individual movements.
Accuracy: Models must be thoroughly tested and made adequate for the tasks of prediction. Models reproducing walking behaviour will very likely present errors when compared to reality and these errors must be known via qualitative and quantitative validations.
Functionality: The majority of the pedestrian models are developed to predict situations where pedestrians are only walking. This is not enough for their application in large facilities where other activities such as waiting, queuing, shopping are also performed.
We point that it is desired that pedestrian models present good computational performance to be of practical use, especially for simulations involving many pedestrians (>1,000 peds simultaneously).
To design walking facilities it is necessary to predict pedestrian behaviours quantitatively and apply it to assess different layouts and plans. The problem can be summarised as: the balance between the cost of construction and maintenance of walking facilities and the different planning objectives.
A common practice is to consider the maximum flow of pedestrians per unity of time (capacity) and apply it to determine corridor widths and size of exits. However, capacity conditions cause the development of queues and congestion and as Fruin already in 1971 recognised: ”capacity design is planned congestion”. Therefore, this is not a good practice since good pedestrian facility should provide safety, efficiency and attractiveness to pedestrians.
Pedestrian models are able to simulate the dynamics of complex walking situations in short periods of time. Furthermore, they allow to quantitatively and objectively compare the performance of different designs. Pedestrian models are developed to cope with complexity of situations and generating visually evidence of the dynamics of situations by means of renderings of the flows.