The Introduction of Adaptive Social Decision-Making in the Mathematical Modelling of Egress Behaviour.
Steve Gwynne
A thesis submitted in partial fulfilment of the requirements of the University of Greenwich for the Degree of Doctor of Philosophy.
Abstract
This dissertation represents an attempt at increasing the behavioural sophistication of evacuation simulations, through the study of evacuation modelling, the development of new behavioural algorithms, their implementation within an existing evacuation model and the testing of the resulting model. This aim is achieved through a number of steps.
Firstly, the range of human behaviour that are exhibited by occupants during the evacuation process is studied. Next, the sophistication of the available evacuation models is investigated and a suitable model is selected and thoroughly assessed (the buildingEXODUS evacuation model). The selected model is then used as a test bed in which to implement the advanced behavioural developments.
The detailed behavioural analysis was conducted to provide the necessary framework, around which an eventual model might be formulated and implemented. This involved the examination of the factors that might influence the occupant’s behaviour, the occupant’s decision-making process and the eventual occupant behaviour.
The mechanisms implemented within the evacuation models presently available were then investigated to determine the current effectiveness of evacuation modelling. This investigation generated possible ideas as to how the modelling process may be conducted and the possible limitations that would be inherent in this process. Rather than creating a completely new behavioural shell, during which time a significant amount of resources would have been diverted into software engineering, an existing behavioural shell was sought after. The buildingEXODUS model was selected as a shell within which the proposed behavioural developments could be analysed for both practical and technical reasons.
The selected model was then validated against a number of experimental and real-life validation cases. This highlighted a variety of limitations and enabled the detailed workings of the selected model to become familiar. In this process, the sophistication and limitation of this shell (the current buildingEXODUS evacuation model) was established. This was required to properly examine the extent of the proposed behavioural development over the existing model.
Once these limitations were established, the proposed developments then had a realistic basis for comparison. The new behavioural features were made in response to sociological, psychological and physical limitations that had been identified in the existing evacuation models. These developments included a more detailed representation of
- The occupant’s familiarity with the enclosure,
- A representation of the occupant’s motivation based on the occupant’s perception of the surrounding conditions,
- Occupant communication,
- Collective behaviour
- And the ability of the occupant to adapt according to the information available.
These proposed behavioural actions and influential factors were then implemented into the buildingEXODUS model. These features were then examined to determine their satisfactory integration into the overall buildingEXODUS model and their impact upon the sensitivity of the model through the use of hypothetical and actual data-sets.
Each of the new behavioural features provided new occupant capabilities and affected the outcome of the buildingEXODUS simulations. The differences may have been centred on qualitative and/or quantitative aspects of the evacuation, depending on the proposed behaviour in question. However, all of the behavioural features examined produced notable results that enhanced the performance of the model in some manner.
Overall the behavioural developments were seen to increase the flexibility and functionality of the model without compromising the previously established ability of the model to cope with the fundamentals of human behaviour. These improvements were therefore seen to further advance the capability of the model to accurately determine the safety of an enclosure during an evacuation through a better understanding of the occupant response and a better and more thorough representation of human behaviour.