Abstract
The desire to visit destinations in order to perform activities is the motivating force for pedestrian movement in public spaces. Understanding how pedestrians decide where to go next is becoming more important as the urban population increases and active modes of transport, such as walking, are seeing widespread adoption in the face of climate change.Use of statistical models is popular for explaining and representing real phenomena. One important step in their development is calibration, where a model is fitted to data.
The first chapter of this thesis describes a simulation study investigating how errors in data collection can affect the subsequent calibration of a destination choice model. Novel methods of adding errors to data and metrics for assessing model calibration are presented. It is shown that errors can introduce bias in parameter estimates, reduce the predictive capability, and alter the resultant dynamics of a calibrated model. The strength of these effects depends on the error type and magnitude.
The second chapter focuses on determining the influence of the environment, how information is presented to individuals, and of any pre-planning on destination choice behaviour, using surveys. Here, the first quantitative results describing their possible effects on calibration are described. Results show that calibration can also be used to define distinct behavioural clusters.
The final chapter identifies a lack of reference data, data which is published openly in a standard format, for pedestrian destination choice model calibration. A reference database is therefore developed via a literature search. Discussions on how the collated data can be used in pedestrian destination choice calibration are provided. The discussion is further illustrated by a case study, using one of the datasets.
This thesis hopefully provides groundwork for further developing pedestrian destination choice models and for future investigations into the complex interactions between data collection methods and statistical model calibration.
| Date of Award | 27 Sept 2022 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | R E Wilson (Supervisor) & Nikolai W F Bode (Supervisor) |
Keywords
- Pedestrian behaviour
- Destination choice
- Statistical modelling
- Model calibration
- Data collection