Planning as Deductive Synthesis in Intuitionistic Linear Logic

We describe a new formalisation in Isabelle/HOL of Intuitionistic Linear Logic and consider the support this provides for constructing plans using deductive synthesis of the proof terms. This representation of plans in linear logic provides a concise account of planning with sensing actions, allows the creation and deletion of objects, and solves the frame problem in an elegant way. Within this setting, we show how planning algorithms are implemented as search strategies within the proof assistant. This allows us to provide a flexible methodology for developing search strategies that is independent of soundness issues. This feature is illustrated in two ways. Firstly, following ideas from logic programming, we show how a significant symmetry in search, caused by context splitting, can be pruned by using a derived inference rule. Secondly, we show how domain specific constraints on synthesis are supported and how they can be used to find contingent or conformant plans.