ARIES Unified Planning Interface
This asset provides the interfacing between Aries and Unified Planning framework developed by The AIPlan4EU project
Aries is an automated planner targeting hierarchical and temporal problems. Aries aims to model and solve hierarchical problems with advanced temporal features and optimization metrics. It relies on these proximities with scheduling problems to propose a compilation into a constraint satisfaction formalism. Solving exploits a custom combinatorial solver that leverages the concept of optional variables in scheduling solvers as well as the clause-learning mechanisms of SAT solvers.
Aries is a project aimed at exploring constraint-based techniques for automated planning and scheduling. It relies on an original implementation of a constraint solver with optional variables and clause learning to which various automated planning problems can be submitted.
A planning problem is then translated into an internal representation based on chronicles: data structures that specify the requirements and effects of an action. Chronicles allow a rich temporal representation of action and are especially useful for representing hierarchical problems where an abstract action can be decomposed into finer-grained ones.
This representation allows for a quite natural encoding of the planning problem into a constraint satisfaction problem that can be solved with our own combinatorial solver.
The current focus of the solver is on hierarchical planning which is especially well suited to represent various robotic and scheduling problems. Non-hierarchical problems are supported but do require more work to reach a state-of-the-art performance (areas of improvement notably include better symmetry-breaking constraints and search heuristics).
Operation Modes Supported
- Oneshot Planning
- Anytime Planning
- Plan Validation
Problem Kinds
- Hierarchical
- Numeric Fluents
- Action Costs
Installation
The library can be easily installed as follows:
python3 -m pip install up-aries