[go: up one dir, main page]

Skip to content

General library for setting up linux-based environments for developing, running, and evaluating planners.

License

Notifications You must be signed in to change notification settings

AI-Planning/planutils

Repository files navigation

planutils

A general library for setting up Linux-based environments for developing, running, and evaluating planners.

There are several ways to use planutils. The easiest ones, based on the provided Docker image, are explained below. If you use planutils for a publication, please cite the following paper:

@InProceedings{muise-et-al-icaps2022systemdemos,
  author       = "Christian Muise and Florian Pommerening and Jendrik Seipp and Michael Katz",
  title        = "Planutils: Bringing Planning to the Masses",
  booktitle    = "{ICAPS} 2022 System Demonstrations",
  year         = "2022"
}

1. Running the latest Docker release

The released Docker image comes with the latest planutils pre-installed. This means that in order to run the latest release, it is not necessary to clone this repository. Note that in order to run some of the planners (all those that are based on singularity), you will need to run Docker with the --privileged option.

Run the planutils container

docker run -it --privileged aiplanning/planutils:latest bash

Active the planutils environment

planutils activate

2. Making your own image with desired solvers

Below is an example for creating your own Dockerfile based on the latest release, with pre-installed solvers.

FROM aiplanning/planutils:latest

# Install solvers and tools
RUN planutils install -y val
RUN planutils install -y planning.domains
RUN planutils install -y popf
RUN planutils install -y optic
RUN planutils install -y smtplan

3. Running planutils from source

You can also run the latest unreleased version. For this, clone this repository and run

docker build . -t planutils-dev:latest

4. Usage

Example of current functionality

$ lama domain.pddl problem.pddl

Package not installed!
  Download & install? [Y/n] y

About to install the following packages: downward (36M), lama (20K)
  Proceed? [Y/n] y
Installing downward...
INFO:    Downloading shub image
 35.88 MiB / 35.88 MiB [=======================================] 100.00% 3.99 MiB/s 8s
Finished installing downward (size: 36M)
Installing lama...
Finished installing lama (size: 20K)
Successfully installed lama!

Original command: lama
  Re-run command? [Y/n] y

Parsing...
$

Example of upcoming functionality

$ planutils install ipc-2018
Installing planners
This will require 3Gb of storage. Proceed? [Y/n]
Fetching all of the planners from IPC-2018 for use on the command line...

$ planutils install server-environment
Setting up a webserver to call the installed planners...

$ planutils install development-environment
Installing common dependencies for building planners...
Installing common planning libraries...

$ planutils install planning-domains
Installing the command-line utilities...
Installing the python library...
Fetching default benchmarks...

$ planutils setup-evaluation configuration.json
Installing Lab...
Configuring Lab...
Ready!
Run eval.py to evaluate
$

5. Add a new package

Package Configuration

  1. Create a folder for the new package, the folder name will be the used to call the planner later.
  2. Set up the install, run, uninstall, and manifest.json files. You can find the template files under packages/TEMPLATE.

Write Manifest file

Create a manifest file named manifest_compact.json if you want to use predefined templates in the packages/TEMPLATE/SERVICE_TEMPLATE folder. The full manifest.json will be generated at run time. You can overwrite the default template by restating the value of some JSON fields.

You can also create a manifest.json file directly if you don't need the template.

Manifest Example

{
    "name": "LAMA-FIRST",
    "description": "http://fast-downward.org/",
    "install-size": "20K",
    "dependencies": [
        "downward"
    ],
    "endpoint": {
        "services": {
            "solve": {
                "args": [
                    {
                        "name": "domain",
                        "type": "file",
                        "description": "domain file"
                    },
                    {
                        "name": "problem",
                        "type": "file",
                        "description": "problem file"
                    }
                ],
                "call": "lama-first {domain} {problem}",
                "return": {
                    "type": "generic",
                    "files": "*plan*"
                }
            }
        }
    }
}

Define Args

There are four types of Args: file, int, string and,categorical. You can add default values for int,string, and categorical arguments.

 "args": [
    {
        "name": "domain",
        "type": "file",
        "description": "domain file"
    },
    {
        "name": "number_of_plans",
        "type": "int",
        "description": "Number of Plans",
        "default":3
    },
    {
        "name": "custom_search_algorithm",
        "type": "string",
        "description": "Custom Search Algorithm",
        "default":"kstar(blind(),k=1)"
    },
    {
        "name": "search_algorithm",
        "type": "categorical",
        "description": "Search Algorithm",
        "choices":[
          {
            "display_value":"Kstar Blind k=1",
            "value":"kstar(blind(),k=1)"
          },
          {
            "display_value":"Kstar Blind k=2",
            "value":"kstar(blind(),k=2)"
          }
        ],
        "default":"kstar(blind(),k=1)"
    }
]

Define Return Types

There are three types of return data: generic, json and log. The generic type should be used for all text based results, the log type should be used for planners like Optic and Tfd which don't generate "proper" plans, and the json type should be used for plans in JSON format.

For the value of files, you will have to write a glob pattern. The planning-as-service backend uses the glob library to find and return all the files that matched.

"return": {
                    "type": "generic/log/json",
                    "files": "*plan*"
                }

About

General library for setting up linux-based environments for developing, running, and evaluating planners.

Resources

License

Stars

Watchers

Forks

Packages

No packages published