Skip to main content

Explainable Multi-Agent Path Finding (XMAPF)

Primary supervisor

Daniel Harabor

Co-supervisors


The multi-agent path finding problem (MAPF) asks us to find a collision-free plan for a team of moving agents. Such problems appear in many application settings (including robotics, logistics, computer games) and a wide variety of solution methods have been proposed. Once a plan is computed, execution proceeds under the supervision of a human operator who is free to modify and adjust the plan, or even reject it entirely, because of changing operational requirements. In these settings we have to be able to “explain” our planning decisions to the operator; for example, why an agent takes a particular path or waits at a particular location. Similarly, we may need to “explain” to the operator why proposed changes are better or worse than an existing solution. In this project we aim to introduce novel methods of generating explanations in MAPF, tackling queries about feasibility, optimality and empirical efficiency of proposed planning solutions.

Required knowledge

- Comfortable with discrete mathematics and proofs

- Basic knowledge of AI (e.g., FIT3080)

- Familiarity with C++

Project funding

Project based scholarship

Learn more about minimum entry requirements.