Output details
11 - Computer Science and Informatics
University of Edinburgh
Exploiting domain knowledge to improve norm synthesis
<22> Originality: The first approach to use abstract state-space search to synthesise norms (behavioural prohibitions that help agents avoid conflict states), while leveraging powerful off-the-shelf planning algorithms.
Significance: The suggested optimisations were shown to reduce the search space from millions of paths to thousands in many domains, as later shown in Christelis' thesis. The kind of abstract planning suggested by the paper inspired our contribution to a multi-million EC-funded project (SmartSociety).
Rigour: Contrary to existing propositional approaches, our method allows first-order theories that capture complex domains much more compactly. AAMAS is the top international conference in the area (24% acceptance rate).