PhD Proposal
Abstract Link to heading
Real-Time Prediction of Spatiotemporal Dynamics in the Built Environment: An AI and Internet of Things Approach Link to heading
Project aim: To develop an AI system for the prediction of spatiotemporal dynamics in the built environment using near real-time geospatial IoT sensor data.
- RO1: Develop tools to assess the quality of near real-time sensor data.
- RO2: Assess the spatiotemporal dependency of near real-time sensor data.
- RO3: Assimilate outputs of agent-based models with real-time sensor data to monitor urban systems in real-time.
- RO4: Evaluate the approach using real-world case-studies and develop a roadmap for scalable deployment.
Predicting short-term behaviour of agents within a city requires reliable high-quality data, real-time processing infrastructure, and models that are capable of generating accurate predictions about complex situations at speed. Enabling this technology could have profound benefits for cities in domains such as emergency response, resource optimisation and decision-making, thus allowing urban areas to better to respond to unsafe situations such as overcrowding or flooding.
Centralised data repositories are appearing in cities around the world, collecting near real-time data from distributed sensor networks. However, there are issues with data quality that need to be addressed before predictive systems can be developed. By applying existing data-quality frameworks to develop a data quality monitoring system this can be resolved (RO1). Making predictions requires an understanding of the dependency of the sensors in time and space – how strongly do sensor measurements relate to the wider sensor network – a spatial AI model in combination with existing knowledge about agent behaviour will be investigated (RO2). To enhance prediction, a data assimilation approach using the outputs of much larger computational models (such as an agent-based model of transport demand) will be investigated (RO3). Ensuring this research has real-world applicability and elevating the technology to a level that meets the needs of its users, a roadmap will be developed. A variety of computational models and sensor network combinations will be investigated to achieve this (RO4).
The research aims to contribute the following to the field of complex systems modelling:
- Greater understanding about the real dynamics of complex urban systems.
- An enhanced understanding of causality in complex urban systems.
And to deliver the following technical capabilities:
- A demonstration of value for the data collected by centralised open urban repositories.
- Pathways to making this data ‘AI ready’.
- A demonstration of a real-time cloud-based web-app that provides useful information derived from the sensor data that can be used for improved decision making.
- A package of code that conforms to best-practice and is built to be compatible with urban digital-twin frameworks such as DAFNI/Gemini.
- A series of publications showcasing any scientific advancements made by this research.