Shaping Collective Traffic Behaviour through Multi-Agent Learning-Based Control
About the Project
The next generation of mobility will be redefined by mixed autonomy — connected autonomous vehicles (CAVs) and human drivers sharing the same road. Each vehicle is an embodied agent: it perceives its surroundings, obeys its own physical dynamics, and must act in real time. Yet the smooth flow and resilience of the whole network emerge from the physical interactions of many such agents, not from any central command. This PhD asks how we can deliberately shape that collective traffic behaviour — from the embodied control of a single autonomous vehicle to the emergent behaviour of the traffic system — using learning-based control and autonomous driving.
Building on advances in deep reinforcement learning, multi-agent systems, autonomous driving and intelligent transportation systems, you will design control and coordination methods — and the mathematical theories behind them — that steer collective traffic dynamics toward efficiency, safety and resilience. Possible directions include embodied perception, decision-making and control for autonomous driving in mixed traffic, multi-agent reinforcement learning for CAV coordination, and the control of mixed human–machine traffic, in which a few learning-enabled vehicles must influence the behaviour of many.
Central questions include: how can a minority of intelligent vehicles influence a largely human-driven system? How does control at the level of a single embodied vehicle — its perception, dynamics and motion — shape the behaviour of the traffic as a whole? And how can learned policies remain safe, robust and trustworthy under complex, real-world road conditions?
You will work with autonomous-driving and traffic simulators — and real-world mobility data where useful — developing scalable algorithms and validating them on realistic scenarios. We seek a candidate with a strong background in AI, machine learning, mathematics, control, engineering, or a related field, with solid programming and mathematical skills and an interest in learning, control and autonomous systems acting in the physical world. Experience with reinforcement learning or optimisation is beneficial but not required.
Informal enquiries, with a CV and short statement of interest, are welcome to Dr Kai-Fung Chu (kaifung.chu@bristol.ac.uk).
Find Your Best Opportunity
Tell them AcademicJobs.com sent you!
