Plan Repair in HTN Planning

During plan execution, a lot can fail! Below is a list of working plan repair approached. Please be aware that they might implement different semantics, in particular with regard to the HTN solution criteria. Keep in mind that plans are only regarded as such in HTN planning if they can be obtained from the initial task network. When plans fail, this requirement might be too strict: Some approaches still demand this, others not.

• IPyHOP, a reentrant version of GTPyho

  • URL: https://github.com/YashBansod/IPyHOP
  • Further info: IPyHOP is a Totally-Ordered Goal Task Network (GTN) Planner written in Python 3.

• Plan Repair via Encoding into HTN Planning:

  • URL: https://github.com/panda-planner-dev/pandaPIpgrRepairVerify
  • Further info: This plan repair approach encodes a given HTN plan repair problem into a standard HTN planning problem. This approach follows the standard HTN solution criteria. This means that despite the plan execution error, final plans still must be a refinement of the initial task network while still taking into account all executed actions before the execution error.

• Stable Plan Repair

  • URL: https://github.com/shop-planner/shop3
  • Further info: The plan repair procedure is part of the SHOP3 system. Further information will be provided soon.

Newer Developments

This entire webpage is basically a private endeavor and therefore usually outdated. Should you know about newer developments, please let me know. One novel repair system (that should still be integrated into the list from above) is from an award-winning ICAPS 2025 paper that compares and evaluates several repair mechanisms.