School of Computer Science and Statistics
Traffic networks can be significantly affected by unexpected events such as accidents, road closures or sudden congestion. Conventional route-planning approaches may struggle to adapt efficiently to these changes. This project investigates whether reinforcement learning can learn adaptive routing strategies that respond to unexpected disruptions. The learned approach will be evaluated in a simulated transport network … Read more
Vehicles approaching highway merges or lane-changing situations need to coordinate their actions to maintain safety and traffic efficiency. This project investigates whether multiple reinforcement learning agents can learn cooperative lane-changing strategies. Different training approaches will be compared with conventional rule-based strategies. The project will evaluate their impact on traffic safety, travel time and overall traffic … Read more
Choosing the best route through a city is challenging because traffic conditions change continuously. Reinforcement learning (RL) provides a way for an agent to learn effective route-selection strategies by interacting with a simulated transport network. This project will investigate whether an RL-based approach can learn to select routes that reduce travel time and congestion compared … Read more
Different travellers may have different preferences when choosing how and when to travel. While some may prioritise travel time, others may prefer cheaper, less crowded or more environmentally friendly options. This project investigates whether reinforcement learning can be used to learn personalised mobility recommendations based on individual preferences and feedback. The project will explore how … Read more
In cooperative multi-agent reinforcement learning, several agents have to learn how to work together. A challenge is that they can sometimes settle on a reasonably good way of cooperating, even though a much better joint solution exists. This project investigates whether the way learning experiences are selected and used during training could influence which solution … Read more
Before fully autonomous driving is achieved, connected autonomous vehicles (CAVs) will operate for a certain period in mixed traffic, which includes both CAVs and human-driven vehicles (HDVs). The dynamic and interactive conditions in mixed traffic scenarios renders CAV decision making particularly challenging. This project will investigate the use of Graph Convolutional Deep Reinforcement Learning for … Read more
This slot is intentionally open. If you have prior hands-on experience with reinforcement learning and an idea you’d like to pursue, whether a novel application domain for RL, or an improvement to an underlying RL algorithm, get in touch to discuss it as a potential project. Prior RL experience is mandatory for this project. This … Read more
Testing software means systematically probing it for the edge cases that break it — but RL policies don’t come with a specification to test against, and their failures often only surface deep into a long trajectory, making them hard to catch before deployment. Counterfactual explanation techniques developed in this group already show that tracing “what … Read more
Balancing competing objectives, eg speed vs. safety, or efficiency vs. fairness, is a core challenge in multi-objective reinforcement learning (MORL). Prior work in my research group has shown how to optimize multiple objectives simultaneously without collapsing them into a single hand-weighted reward. But specifying the right trade-off between objectives numerically is hard: people often can’t … Read more
Recent work in my research group has shown that AI feedback can help balance competing objectives, like vehicle delay, emissions, and fairness across approaches, in traffic signal control, using preferences over individual decisions to guide reinforcement learning. However, decision-making in traffic control in sequential and long-horizon; a signal-timing choice that is suboptimal in the short-term … Read more