An Efficient Binding Algorithm for Forward-chaining HTN Planning Based on Object-oriented Knowledge

An efficient improvement on the binding process of SHOP2 planner for hierarchical task network (HTN) planning is proposed by using some object-oriented formal knowledge. Two object-oriented structures, Structure of Object and Structure of Object Set are defined to separate SHOP2psilas world state to two parts, object-oriented knowledge and predicate-based atoms. By preprocessing, improved binding algorithm can avoid searching in predicate-based state atoms when reasons about the type of object and the membership relation between two hierarchical objects, and thus the CPU time cost can be reduced. Experiments show that this approach can significantly increase the efficiency of binding process for forward-chaining HTN planning in systems concerning a large number of objects and with hierarchical structures without losing generality and easiness in modeling process.