A formalization of multi-agent planning with explicit agent representation

What is it about?

Our work presents a formalization of multi-agent planning problems where agents are explicitly represented. Unlike previous formalizations, we discuss the advantages of explicitly representing agents and show the implementation in the Unified Planning formalism and some practical examples. The proposed formalism is equivalent to other models, particularly the Multi-Agent Planning Domain Definition Language (MA-PDDL). It can therefore be compiled into it and solved by existing multi-agent planning solvers. Additionally, we introduce a further extension to define more complex multi-agent problems with explicit models of other agents. Our work aims to simplify the modeling and solving of multi-agent planning problems, enhancing the clarity and flexibility of domain and problem specifications. The main goal is to clearly represent multi-agent planning problems, reducing possible causes of semantic confusion. The proposed formalization allows compiling multi-agent problems into MA-PDDL and utilizing available multi-agent planning solvers.

Featured Image

Photo by Gerard Siderius on Unsplash

Photo by Gerard Siderius on Unsplash

Why is it important?

We present a formalization of multi-agent planning problems with explicit agent representation. They highlight the difference between agents and objects, emphasizing that agents are actively operating entities in the environment while objects are passive. The formal model proposed in this paper allows for clear semantics and reduces possible semantic confusion compared to existing models like Multi-Agent Planning Domain Definition Language (MA-PDDL). Furthermore, this formalism is implemented in the AIPlan4EU - Unified Planning Framework, making it easy and efficient to create and edit multi-agent problems, compile them into MA-PDDL, and solve them with multi-agent planners.

Perspectives