Available Engines

The tables below give a high-level overview of the planning engines integrated with the UP.

The characterization of the planning systems is deliberately kept very broad and is mainly intended to help identify which planners are relevant for a given class of problem. We redirect you to the specific documentation of each planner for a more in depth characterization of their features and limitations.

In this page only appear solvers that are officially integrated in the unified planning library, for which we have reasonable confidence that they will not give you incorrect results (i.e. they successfully pass all engine tests), now and in the future (i.e. they have a clearly identified maintainer). We do mention some solvers whose integration is partial, but that may be nevertheless of interest in some special cases.

Action-Based Planning

Engine

Operation modes

Classical

Numeric

Temporal

Metrics

Fast-Downward

OneShot, Anytime

Y

plan length, action costs

ENHSP

OneShot, Anytime

Y

Y

action costs, final value

Tamer

OneShot

Y

Y

Y

LPG

OneShot, Repair, Anytime

Y

Y

Y

plan length, makespan, action costs

Aries [1]

OneShot, Anytime

Y

Y (integers)

Y

plan length, makespan, action costs

Pyperplan [2]

OneShot

Y

Plan Validation

Engine

Action-Based planning

Numeric

Temporal

Hierarchical

Scheduling

Multi-Agent

UP (builtin)

Y

Y

Tamer

Y

Y

Y

Aries

Y

Y

Y

Y

Y

MA-Plan-Validator

Y

Y

Hierarchical Planning

Engine

Operation modes

Total-Order

Partial-Order

Numeric

Temporal

Metrics

Aries

OneShot, Anytime

Y

Y (integers)

Y

Y

plan length, makespan, action costs

A WIP integration is known for SIADEX.

Scheduling

The only planner with full support for scheduling is Aries. Integration work is known for the discrete-optimization suite and for PPS.

Multi-Agent Planning

Engine

Operation modes

Classical

Partial-Order

fmap

OneShot

Y

Y

Task and Motion Planning

Support for task and motion planning is currently provided by the Spiderplan engine.