ASET: a Multi-Agent Planning Language with Nondeterministic Durative Tasks for BDD-Based Fault Tolerant Planning

In this paper, we introduce a multi-agent planning language called ASynchronous Evolving Tasks (ASET). The main contribution of ASET is a novel explicit representation of temporally extended tasks that may be nondeterministic both with respect to duration and effects. Moreover, ASET explicitly models the environment as a set of uncontrollable agents. We formally define ASET descriptions and their transformation to a nondeterministic planning domain. Using a Boolean encoding, fault tolerant planning problems specified in ASET can be solved efficiently with state-of-the-art BDD-based planning systems. Our preliminary experimental results show that the transformation of ASET domains to nondeterministic planning domains is computationally efficient even for ASET descriptions with a high level of temporal detail.