PANDA Planning System

The PANDA planning system allows to solve different kinds of planning problems. The planning algorithm for all these problems is a hybrid planning approach, which fuses hierarchical planning with causal reasoning. It can solve the following classes of planning problems:

  • classical, e.g., as defined by PDDL, and partial-order causal-link (POCL) planning problems,
  • hierarchical task network (HTN) problems, and
  • hybrid planning problems

We are currently in the process of providing our code as Open Source. We will provide it here when we are done.

Recommended literature directly related to the planning system:

From Abstract Crisis to Concrete Relief (A Preliminary Report on Combining State Abstraction and HTN Planning) by Biundo und Schattenberg (ECP 2001) contains:

  • A formalization of hybrid planning based on predicate logic
  • A formalization (based on predicate logic) of the legality of decomposition methods: Under which circumstances is a method a correct implementation of its abstract task?

Hybrid Planning Heuristics Based on Task Decomposition Graphs by Bercher et al. (SoCS 2014) contains:

  • How do hybrid planning models and problems look like?
  • What are the solution criterio of hybrid planning problems?
  • Planning algorithm: How does the search work?
  • The heuristic MME, which is specifically designed for hybrid planning. It estimates the number of modifications that are required to reach a goal. It does so by incorporating the task decomposition graph, which represents the domain's task hierarchy.

More than a Name? On Implications of Preconditions and Effects of Compound HTN Planning Tasks by Bercher et al. (ECAI 2016) contains:

  • Complexity results for hybrid planning: How hard is it so solve hybrid planning problems? How hard is it to tell whether a potential solution is actualy a solution?
  • An overview of formalizations from the literature that also fuse hierarchical planning with POCL planning.
  • An overview of legality criteria from the literature: Under which circumstances is a method a correct implementation of its abstract task?

An Admissible HTN Planning Heuristic by Bercher et al. (IJCAI 2017) contains:

  • An admissible (cost-sensitive) heuristic for hierarchical and hybrid planning. It thus allows to find optimal solutions. In the same way than the MME heuristic does, this one takes the hierarchical structure of abstract tasks into account - based on the task decomposition graph.

Download of the Planning System

  • We offer our planning under an open source licence. The download can be found on our PANDA overview page.