Ivana Dusparic

I am open to supervising projects developing novel artificial intelligence techniques and/or applying these techniques in robotics, intelligent urban mobility and smart cities in general. In particular, I am interested in learning-based agents and multi-agent systems, with particular focus on reinforcement learning (RL), including deep reinforcement learning, transfer learning, lifelong learning, credit assignment, multi-agent collaboration, self-adaptive systems, and self-aware systems . I am also interested in applying these techniques to robotics as well as emerging urban transport models and their impact on cities, eg intelligent urban traffic control. I am also open to other proposals applying learning-based multi-agent techniques to management and optimization of other large-scale infrastructures - if you have an interest in learning-based optimization and have an application area in mind, let me know!

TAKEN – Propose Your Own: RL Applications or Algorithmic Improvements

Ivana Dusparic (email | all projects)

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

TAKEN – Test-Driven RL: Automated Failure-Case Discovery for Trained Policies

Ivana Dusparic (email | all projects)

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

TAKEN – Human-Feedback Fine-Tuning for Multi-Objective RL

Ivana Dusparic (email | all projects)

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

TAKEN – Trajectory-Level Preference Learning for Traffic Signal Control

Ivana Dusparic (email | all projects)

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

TAKEN – Building a Multi-Agent Negotiation System for Traffic Signals

Ivana Dusparic (email | all projects)

This idea will explore traffic signal control done by AI agents that negotiate in natural language: a vehicle requesting priority, a traffic light balancing competing demands, a neighbouring intersection protecting its own green-wave timing. This project builds the foundational system: a working prototype combining a traffic simulator (SUMO) with large language model agents that negotiate … Read more