A Survey on Hierarchical Planning – One Abstract Idea, Many Concrete Realizations

Abstract

Hierarchical planning has attracted renewed interest in the last couple of years, which led to numerous novel formalisms, problem classes, and theoretical investigations. Yet it is important to differentiate between the various formalisms and problem classes, since they show -- sometimes fundamental -- differences with regard to their expressivity and computational complexity: Some of them can be regarded equivalent to non-hierarchical formalisms while others are clearly more expressive. We survey the most important hierarchical problem classes and explain their differences and similarities. We furthermore give pointers to some of the best-known planning systems capable of solving the respective problem classes.

Delete- and Ordering-Relaxation Heuristics for HTN Planning

“…Höller et al [2018] introduced an approach that relaxes the HTN planning problem to a classical planning problem. In contrast to the translation-based approaches by Alford et al [2009;2016] that do not change the set of solutions and use a classical planner for search, the translation of Höller et al increases the set of solutions, i.e. it relaxes the problem.…”

“…Interestingly, what is encoded in the classical model can be seen as the building process of a DT [Höller et al, 2019]. Bit-Monnot et al [2016] introduced a translation from HTN-like models into a temporal (non-hierarchical) planning model that is used for a reachability analysis during search, which prunes wide parts of the search space and is combined with blind search (so technically, no heuristic is used). The translation marks via state features which abstract tasks are currently processed, allowing only the insertion of tasks belonging to methods that decompose these tasks; this minimizes task insertion.…”

“…However, since SAT-based approaches currently perform best, we also include those into the evaluation. RC ADD , RC FF , and RC LMC use progression search with heuristics based on a relaxation by Höller et al [2018;2020] combined with the Add [Bonet and Geffner, 2001], FF [Hoffmann and Nebel, 2001], and LM-Cut [Helmert and Domshlak, 2009] heuristics from classical planning, TDG c and TDG m use plan space search with the TDG-based heuristics introduced by , and UMCP is the search strategy by Erol et al [1994] implemented in the PANDA system. SAT'18 and SAT'19 are translations into propositional logic as introduced by Behnke et al [2018; 2019].…”

HDDL: An Extension to PDDL for Expressing Hierarchical Planning Problems

“…HTN planning with task insertion. An overview of hierarchical problem classes can be found in the survey by Bercher, Alford, and Höller (2019). Some problem classes are even syntactically equivalent to HTN planning and only differ in their solution criteria.…”

“…Much progress has recently been made in hierarchical planning (Bercher, Alford, and Höller 2019). Novel systems based on the traditional, search-based techniques have been introduced (Bit-Monnot, Smith, and Do 2016; Shivashankar, Alford, and Aha 2017;Bercher et al 2017;Höller et al 2018;, but also new techniques like the translation to STRIPS/ADL (Alford, Kuter, and Nau 2009;Alford et al 2016a).…”