The article Decentralized Planning Using Probabilistic Hyperproperties by Francesco Pontiggia, Filip Macák, Roman Andriushchenko, Michele Chiari, and Milan Češka, introduces an innovative approach to decentralized planning of multi-agent systems. This is achieved by extending classical probabilistic logic with hyperproperties, which allow for effective reasoning about sets of runs. This method enables to capture complex temporal requirements for the behaviour of multiple agents while also expanding the possibilities for planning such systems compared to traditional methods.
The work presents significant theoretical results, including a generalization of undecidability for decentralized planning and a practical approximate algorithm that surpasses existing solutions by providing higher-quality planning strategies. The article was accepted to the prestigious 24th International Conference on Autonomous Agents and Multiagent Systems (AAMAS'25), which is categorized in the highest tier, CORE A*.
This year's conference received over 1000 submissions, with less than 25% accepted for publication. The AAMAS'25 jury awarded the paper the Best Student Paper Award, one of only three awards announced at the conference. The research was conducted under the supervision of Associate Professor Milan Češka, Ph.D., in collaboration with his doctoral students from Brno University of Technology and two colleagues from TU Wien, as part of the VASSAL project.
