ryanxwang / up-enhsp

This is the integrator for ENHSP, the Expressive Numeric Heuristic Search Planner. ENHSP is a planner supporting classical and numeric state variables, and is able to compute both satisificing and optimal plans in a one-shoot or anytime manner.

ENHSP Unified Planning integrator

This is the ENHSP UP integrator. ENHSP is an expressive numeric planner supporting planning problems involving Boolean and numeric state variables, actions, processes and events as those that can be expressed using the PDDL+ language. A distinctive feature of ENHSP is the ability to reason over problems with a prevalent numeric structure, which may involve linear and non-linear numeric conditions. ENHSP favors the capability of modeling expressive problems; indeed it handles disjunctive preconditions and conditional effects without compiling them away through potentially expensive compilations (e.g., exponential conversion into DNF). ENHSP can be used as a satisficing or optimal planner, yet optimality is only guaranteed for fragments of PDDL+, in particular so called simple numeric planning problems. ENHSP is written completely in JAVA. ENHSP has been firstly presented here, yet it contains many features studied in several other papers from the AI Planning literature.

Installation

After cloning this repository run pip install up-enhsp. up-enhsp can also be installed through the unified-planning framework with the command pip install unified-planning[enhsp]. ENHSP will be downloaded and compiled directory from the public repository of ENHSP. ENHSP, and therefore this integrator requires the openjdk 17 installed on your machine.

Planning approaches of UP supported

Numeric Planning

Default configuration

Greedy Best First Search as search algorithm + the hadd heuristic.

Operative modes of UP currently supported

One shot planning

Acknowledgments

This library is being developed for the AIPlan4EU H2020 project (https://aiplan4eu-project.eu) that is funded by the European Commission under grant agreement number 101016442.