Modelling Goal-Oriented Pedestrian Behaviour
Lead Research Organisation:
University of Bristol
Department Name: Engineering Mathematics
Abstract
Developing intelligent cities has become a major economic activity in the 21st century and an unprecedented number of people will live together in urban areas all over the globe. This represents incredible challenges in the field of transport. Rising urbanisation and mobility lead to busier transportation hubs, shopping malls and public spaces that need to be carefully designed and managed to cope with growing demand.
This project will involve working on computer simulation models for pedestrian crowds that are currently revolutionising the industry. Pedestrian movement is self-organised, as individuals respond to the movement of others whilst walking. Built on mathematical representations of individual behaviour, simulation models are a versatile tool that promise to accurately predict the flow patterns observed in real crowds. The project will address a crucial open problem by developing a model for how individuals plan, review and execute their activities when navigating through any public space. The intended outcome of this project is a general model that can accept building configuration and individual objectives as inputs and accurately predict crowd behaviour under a variety of situations. These situations will include evacuations in the event of fires or other hazards, and times of high thoroughfare.
There is also potential to develop and carry out real-world experiments to refine the model. This could be done by using lab-based trials with groups of volunteers in carefully designed environments. This will require suitable settings to conduct the experiments, advertising/canvasing for volunteers and possible incentives for participation. Another potential experimental framework is based around an 'educational video game' which could be distributed in public areas where volunteers complete objectives in the video game. Data would then be collected from participant decisions within the game and from their feedback. The game would require careful planning and designing to keep users interested while adhering to the needs of the project.
This project will involve working on computer simulation models for pedestrian crowds that are currently revolutionising the industry. Pedestrian movement is self-organised, as individuals respond to the movement of others whilst walking. Built on mathematical representations of individual behaviour, simulation models are a versatile tool that promise to accurately predict the flow patterns observed in real crowds. The project will address a crucial open problem by developing a model for how individuals plan, review and execute their activities when navigating through any public space. The intended outcome of this project is a general model that can accept building configuration and individual objectives as inputs and accurately predict crowd behaviour under a variety of situations. These situations will include evacuations in the event of fires or other hazards, and times of high thoroughfare.
There is also potential to develop and carry out real-world experiments to refine the model. This could be done by using lab-based trials with groups of volunteers in carefully designed environments. This will require suitable settings to conduct the experiments, advertising/canvasing for volunteers and possible incentives for participation. Another potential experimental framework is based around an 'educational video game' which could be distributed in public areas where volunteers complete objectives in the video game. Data would then be collected from participant decisions within the game and from their feedback. The game would require careful planning and designing to keep users interested while adhering to the needs of the project.
People |
ORCID iD |
| Richard Edward Wilson (Primary Supervisor) | |
| Christopher King (Student) |
Publications
King C
(2023)
Towards a Reference Database for Pedestrian Destination Choice Model Development
in Collective Dynamics
Tong Y
(2021)
Using agent-based simulation to assess disease prevention measures during pandemics*
in Chinese Physics B
Studentship Projects
| Project Reference | Relationship | Related To | Start | End | Student Name |
|---|---|---|---|---|---|
| EP/N509619/1 | 30/09/2016 | 29/09/2021 | |||
| 2120427 | Studentship | EP/N509619/1 | 30/09/2018 | 30/03/2022 | Christopher King |
| EP/R513179/1 | 30/09/2018 | 29/09/2023 | |||
| 2120427 | Studentship | EP/R513179/1 | 30/09/2018 | 30/03/2022 | Christopher King |
| Description | A novel method of adding errors to artificial data that are controlled, reproducible, and make the data more realistic. Also, new metrics for assessing how well a model fits data have been defined. One result is that errors in data can reduce the impact of factors which the data describes. Another result is that, depending on the original behaviour being modelled, errors can significantly alter the behaviour displayed by a fitted model. The type of error applied has an impact on these two results. Preliminary results from an online survey of destination choice in virtual environments show that how information on the environment is presented can significantly influence a person's decisions. Also, there is evidence that people plan the series of destinations in advance of arrival and that this plan can be either developed by the person or provided from an external source. An open-access reference database is produced from contemporary literature comprised of a variety of datasets. An illustration of how this data could be used in pedestrian destination choice model development is shown. This, and a discussion of the datasets, highlights general issues when publishing reference data for subsequent model development. |
| Exploitation Route | The novel method of adding errors and the model fitting metrics could be used by any researcher working with discrete, continuous, or sequential data to fit a statistical model. The methods described can be applied to many different pedestrian destination choice models in the future. If the errors in real data can be estimated, then the predicted impact of the errors on the subsequent model fitting can be supposed and used to assess the viability for the model in future uses. The online surveys can be expanded to look at how the environment configuration and information presentation on the decision behaviour of people in more detail. The reference database created here will hopefully be used by other researchers, both within and without pedestrian destination choice, in any number of potential applications and/or model development. Hopefully this will encourage researchers in pedestrian destination choice behaviour and beyond to publish any collected data openly, therefore, this database will hopefully grow. |
| Sectors | Transport |
| Title | Agent-based simulator of pedestrian behaviour |
| Description | Simple tool which can simulate the movement, routing, and destination choice behaviour of individual agents in any given environment. Can generate trajectories of each agent over time, the number of agents at different places in the environment, as well as distances at which each agent is from particular areas. |
| Type Of Material | Computer model/algorithm |
| Year Produced | 2020 |
| Provided To Others? | No |
| Impact | Simple, easily-modifiable model for agent-based simulations in pedestrian destination choice behaviour. Could be useful for future members in the research group in this field. |
| Title | Synthetic error generation |
| Description | Method for adding realistic measurement errors to simulated data from agent-based models. Errors in chosen destination sequences, and measured occupancies and distances can be implemented in systematic and controlled ways. Sequence errors take the form of substitutions, additions, and/or deletions. Occupancy and distance errors are sampled from a distribution based on the true value. |
| Type Of Material | Data analysis technique |
| Year Produced | 2020 |
| Provided To Others? | No |
| Impact | Generic methods for adding synthetic measurement error to simulated data. This can improve the realism of generated data by producing realistic errors in controlled ways. Allows the impact of such errors on subsequent data analysis to be investigated. |
| Description | Engineering Postgraduate seminar |
| Form Of Engagement Activity | A talk or presentation |
| Part Of Official Scheme? | No |
| Geographic Reach | Local |
| Primary Audience | Postgraduate students |
| Results and Impact | 40-minute presentation of recent research to other postgraduate research students within the department (~10), sparking questions and discussions in both cases. This helps keep my peers up-to-date with my research and allows the opportunity to obtain new ideas from outside my research discipline. |
| Year(s) Of Engagement Activity | 2020,2021 |
| Description | Pedestrian Dynamics Postgraduate Symposium |
| Form Of Engagement Activity | A talk or presentation |
| Part Of Official Scheme? | No |
| Geographic Reach | International |
| Primary Audience | Postgraduate students |
| Results and Impact | Gave a talk about some ideas for future research regarding the open publication of data in pedestrian destination choice for discussion with other postgraduate students in pedestrian dynamics. This provoked questions and discussion, with some suggestions which were later used in the research. Raised awareness of the lack of open data and encouraged attendees to publish any collected data open in the future. |
| Year(s) Of Engagement Activity | 2021 |
| Description | Pedestrian Dynamics workshop |
| Form Of Engagement Activity | A talk or presentation |
| Part Of Official Scheme? | No |
| Geographic Reach | International |
| Primary Audience | Postgraduate students |
| Results and Impact | Gave a talk of previous research and results to postgraduate research students in pedestrian dynamics from Germany, the UK, the Netherlands, and other central European countries. This sparked questions and discussion afterwards and raised awareness of my research interests to the wider pedestrian dynamics community. |
| Year(s) Of Engagement Activity | 2020 |
| Description | School Postgraduate Conference |
| Form Of Engagement Activity | A talk or presentation |
| Part Of Official Scheme? | No |
| Geographic Reach | Local |
| Primary Audience | Postgraduate students |
| Results and Impact | Internal online conference for postgraduate researchers within the School to communicate their research to fellow postgraduates within the School and to learn about the work of others. Gave a 10-minute talk on some of my recent work with 2-minute Q&A from the audience. Raised awareness of my research to the postgraduate community within the School and practised communicating my research to others. |
| Year(s) Of Engagement Activity | 2020,2021 |