# The TOAD Planning System

TOAD is a planning system for Totally Ordered (TO) HTN planning. It translates the problem to a classical planning problem. For a subclass of TO problems, this is possible without relaxation, e.g., the solution found for the generated classical problem is guaranteed to be a solution for the HTN planning problem. Due to the differences in expressivity, this is not always possible. When it is not, the problem is transformed into a classical planning problem that has a superset of solutions, i.e. the plan found by the classical problem is not necessarily a solution for the HTN planning problem. TOAD provides a verification mechanism to check whether it is a plan or not.

You can download the source code here.

To use the planner, perform the following steps:

- Ground the problem using the PANDA grounder. Let ground.panda be the file with the grounded model.
- Call TOAD: toad ground.panda. TOAD generates a classical planning problem. It uses the sas+ format used by Fast Downward as standard output format. Let problem.sas be that file.
- Use Fast Downward with your favorit configuration to solve that problem, e.g.: ./downward --evaluator 'hff=ff()' --search 'lazy_greedy([hff], preferred=[hff])' --internal-plan-file sas_plan < problem.sas. Let sas_plan be the generated solution.
- From the output of step 2 it can be seen whether the translation used an approximation. If it did, perform the following steps to verify whether the solution is also one for the HTN planning problem.
- Generate a verification problem using toad -v ground.panda sas_plan
- The output of this process is a usual HTN planning problem. It has a solution if and only if the solution is a solution for the HTN planning problem. We used the progression search of the HTN planning system PANDA to solve it. Usually, a configuration using DFS will solve it (use a heuristic search if this fails).

Please cite the following when referencing TOAD:

- D. Höller.
**Translating Totally Ordered HTN Planning Problems to Classical Planning Problems Using Regular Approximation of Context-Free Languages**. In: Proceedings of the 31st International Conference on Automated Planning and Scheduling (ICAPS). AAAI Press, 2021.