The ultimate aim of my research is to build explainable sequential decision-making systems.

 

By explainable I mean human-centric AI agents that are understood by their users, that understand their users, and that are effective and trustworthy collaborators for their users. By sequential decision-making I mean tasks that extend over time, and require an autonomous agent to make many smart choices to achieve some user goal.

To achieve this, my research focuses on planning, required for acting towards long-term goals; on learning, required for acting in unknown environments; and on the explainability of planning and learning, required for successful human-AI collaboration.

In order to tackle these challenges, I often use a variety of traditional and digital games as testbeds. I then transfer game AI technology to non-game domains in collaborative projects with industry, in sectors such as logistics and transportation, smart manufacturing, and sustainable energy.



 

Current projects I am involved with:

PEER

Improving human-AI collaboration on complex planning and optimization tasks, with several applications from manufacturing, to retail, to making cities more inclusive. [Leading the work package on AI]

 

Align4Energy

Multi-agent, multi-objective learning and planning, with the goal of building a decision support system for policy makers and citizens that facilitates more sustainable energy use. [Co-leading the work package on AI]

Dynaplex
Deep reinforcement learning and planning with applications in logistics, such as transportation management, inventory management, and warehouse management.


 

Current PhD candidates I am (co-)supervising:

Luca Begnardi

Gao Peng

Bart von Meijenfeldt

Chengpeng Hu

Athanasios Trantas


 

Here are a few videos describing some of my past research: