Below is a list of potential PhD projects available within our research group. Please get in touch with the relevant project supervisor to discuss.
- Driverless cars and systems-level effects
- Using virtual experiments to study crowd evacuations
- Modal Decomposition of delay coupled networks
- Modelling human mobility flows worldwide using public online geographical data
- Opinion diffusion logics for swarms
- Modelling and engineering cell dynamics in intestinal organoids
- Glassy dynamics in animal populations
- Exploring patterns of vehicle ownership and use
- Engineering collective dynamics in mammalian cells using microfluidics
- Criticality of the inner ear
- Distributed learning of imprecise probabilities in robot swarms
- Analysis of social interaction patterns in healthcare contexts
- Individual-centric data analytics in pedestrian crowds
Google (and others) have recently made great strides towards developing an autonomous car that can drive safely, with very little human input. But so far, the question of how many autonomous vehicles will interact with each other and with human drivers is largely neglected. Your work will help predict how travel time, capacity and emissions will change in this future autonomous-human hybrid system. Fundamental questions include whether autonomous vehicles should (if it were possible) drive exactly like humans or indeed whether they should drive non-deterministically to avoid being gamed. Depending on your background, there are a variety of possible ways into this project, including analytical techniques and simulations with/without real humans in the loop.
Crowd evacuations arise when large numbers of people exit a building or vehicle during an emergency. Understanding the dynamics of these evacuations will help the design of safe facilities and the development of risk mitigation strategies. Recreating realistic crowd evacuations in experiments is expensive, time-consuming and potentially dangerous. As an alternative starting point, this project will use virtual experiments to test the behaviour of evacuees by integrating avatars controlled by people into computer-simulated crowds. The goal of this research is to develop highly refined mathematical models to predict peoples’ behaviour with a focus on heuristics for route choice, activity scheduling and information use.
In many agent based systems the model of the system can be described by identical agents that are coupled through delayed signals. Such system can be a network of autonomous vehicles connected through wireless communication. In order to better design such systems it is desirable to understand what modes of operations are the dominant and what is the dynamics within that mode. As simple linear decomposition method was already proposed in R. Szalai and G Orosz, Phys. Rev. E 88, 040902, 2013. The task within this project is to extend this theory to nonlinear network models.
Mathematical models to estimate individual and collective human mobility have important applications, including transportation, urban planning and epidemic modelling. The aim of this project is to create a statistical model to estimate commute and migration flows worldwide using as input detailed geographical information publicly available online. The first part of the project consists in the creation of a spatial database storing high-resolution geographical information, such as land use, points of interests, transportation network. Data will be primarily extracted from Openstreetmap and online social networks. In the second part of the project such data will be fed to a neural network model that will be trained to estimate human mobility flows, specifically commute and migration flows provided by census surveys. The expected advantage of this approach with respect to other state of the art mobility models is the greater flexibility to combine raw geographical data in order to extract complex relevant features allowing to achieve higher accuracy and generalisation power in its predictions.
A fundamental challenge in swarm robotics is that of how best to propagate information across the population on the basic of only local interactions between robots. In a complex noisy environment it may be more appropriate to represent knowledge qualitatively rather than quantitatively. For example, in comparing different possible decisions, actions or policies it may be sufficient to identify a partial ordering of these rather than attempting to quantify the value of each. This project will investigate possible approaches to opinion diffusion in robotic swarms in which individual agents’ beliefs take the form of logical formulas rather than being represented by numerical values. More specifically, we will study how such opinions propagate under a range of belief combination models. The project will involve a mixture of simulation and Kilobot experiments in which robots will hold qualitative beliefs in the form of canonical logical formula. There will also be an opportunity for theoretical studies using a combination of formal logic and network theory.
Organoid cultures have been recently proposed as a valuable tool to study in vitro tumour progression, invasion and drug response. Their three-dimensional structure provides a good physiological representation of tissue organization, functional differentiation and response to chemical and mechanical signals. The complex and multiscale nature of organoids requires mathematical modelling to gain quantitative understanding of experimental measurements and predict response to external inputs (e.g. cell-cycle modulators). Building on an existing 2D agent-based model 1 (see figure), a PhD project in this area would be focussed on developing intestinal organoid 3D multiscale models capable of simulating phenotypic changes and intercellular communication, fitting experimental data generated by collaborators in the School of Cellular and Molecular Medicine. Model predictions about the dynamic response to drugs could be tested within an available microfluidics/microscopy platform.
The collective dynamics of animals has been a hot topic for scientists across a plethora of disciplines spanning biology, computer science, applied mathematics and physics due to its multiple roles in understanding fundamental questions in biology and on complex systems as well as its potential to develop bio-inspired robotic applications. Recent findings on animal collective studies have open up another gap-bridging area of research connecting material science and biology. Under rather broad conditions populations of scent-marking animals possess in fact all salient features of glassy materials. Animal collective movement models may thus serve as a further gap-bridging way to help understand how metastability emerges in complex systems as well as to answer fundamental questions about animal ecology.
It is estimated that 40,000+ people in the UK die prematurely each year due to air quality problems caused predominantly by diesel cars, buses and lorries. I have recently shown how the government's licensing and MOT test data can be fused to understand where different kinds of vehicles are kept and how much they drive. In this project, you will develop methods that determine *where* those vehicles drive. Depending on your background, you might attempt a data science approach which fuses other sources of mobility data that we are developing, including number plate recognition and bluetooth/wireless sniffing; or an analytical/simulation approach which uses trip models to predict flows at a link level. This is a pressing policy problem and you will have the opportunity to interact with stakeholders at Bristol City Council and the central government Department for Transport.
We are currently developing computational and experimental tools to externally control signalling pathways and cell-cycle dynamics in populations of embryonic stem cells. External feedback (see figure) is implemented in real-time using a microfluidic/microscopy platform, which allows precise control and dynamic perturbation of the local microenvironment of living cells whilst measuring the cellular response (i.e. fluorescent proteins, cell morphology) 2 . A PhD project in this area could focus on one or more of the following topics: (1) Developing robust segmentation and control algorithms, (2) Single-cell control; (3) Automatic reprogramming and differentiation of somatic and pluripotent cells, respectively; (4) Design and implementation of novel microfluidics devices to host different cell types. Students would be involved in both the computational and experimental aspects of this project.
The inner ear is a highly tuned organ. It is sensitive to sound pressure levels down to atmospheric noise. In order to reach this ideal sensitivity the ear need to be tuned. The central research question of this project is how the self-tuning is achieved by the cochlea, which is expected be an example of self-organised criticality. The research can utilise the results of previous work that, which has produced an accurate nonlinear model of the inner ear.
Distributed belief updating is a key part of decision making in multi-agent autonomous systems. In the presence of uncertainty individual beliefs are often represented as probability distributions over a set of hypotheses of interest. Updating can then be either Bayesian or can employ probability pooling in which agents adapt their beliefs based on those of their neighbours. In either case, updated beliefs continue to take the form of precise probabilities. However, in the face of conflicting probabilistic information it maymore representative to take beliefs as corresponding to convex sets of probability distributions i.e. to imprecise probabilities. This project will explore distributed belief updating for imprecise probabilities. Agent-based simulation experiments will be conducted using different updating methods to explore their system-level properties. There will then also be an opportunity to investigate these systems analytically. Finally, we will apply this approach to the best-of- n decision problem for robot swarms using the kilobot platform.
Social interactions, the processes by which we react to and influence those around us, are fundamental to human nature. Increasingly, evidence suggests that ageing, mental well-being or physical disease impact on the patterns of peoples’ social encounters and vice-versa. Digital health augments diagnosis and treatment in healthcare based on analysing automatically recorded physiological, behavioural or biological data. Currently, digital health emphasises data from individuals. In contrast, this project will focus on demonstrating how sensor arrays for automated data acquisition in combination with novel computational analysis of behavioural signatures can be used to enhance healthcare. The specific use-case for this project will be interactions in patient-doctor consultations and we will aim to lay the foundations for directly applicable digital healthcare support systems.
Cities across the world are full of pedestrians who need facilities to move comfortably, efficiently and safely through the urban landscape. This project will use the latest technology in computation, data analytics and mathematical modelling to transform our understanding of pedestrian crowd movement. Currently our understanding of pedestrian crowd movements is informed by data derived from a top-down view of pedestrian facilities. Instead, we will take the perspective of individuals when collecting data by mounting cameras and additional sensors directly on each individual in experiments with crowds. Using this unprecedented data we will be able to answer question on where within crowds external visual cues can be perceived and on the extent to which people within crowds can perceive the movement of the entire crowd, for example.