SFPlanner

Author: Herry (herry13@gmail.com)
License: BSD

A Ruby script and library of SFP planner, which solves a planning task written in SFP language.

Click here, for more details about SFP language.

Features

Automatically generating a sequential or partial-order (parallel) plan.
Use JSON as the standard format of the plan.
Generate an image file that illustrates the graph of the plan.

To install

$ gem install sfplanner

Requirements

Ruby (>= 1.9.2)
Graphviz
Rubygems
- sfp

Tested on:

Ubuntu 12.04
Debian Squeeze
Scientific Linux 6
MacOS X 10.8

To use as a command line

solve a planning task, and then print the output in JSON
```
  $ sfplanner <sfp-task-file>
```

To use as Ruby library

parse an SFP file, and then generate the plan (if found) in Hash:

 # include sfplanner library
 require 'sfplanner'

 # solve and return the plan in Hash
 planner.solve({:file => file_path})

parse an SFP file, and then generate the plan in JSON:

 # include sfplanner library
 require 'sfplanner'

 # solve and return the plan in Hash
 planner.solve({:file => file_path, :json => true})

Example of Planning Task

Create file types.sfp to hold required schemas:

 schema Service {
 	running is false
 	procedure start {
 		conditions {
 			this.running is false
 		}
 		effects {
 			this.running is true
 		}
 	}
 	procedure stop {
 		conditions {
 			this.running is true
 		}
 		effects {
 			this.running is false
 		}
 	}
 }
 schema Client {
 	refer isref Service
 	procedure redirect(s isref Service) {
 		conditions { }
 		effects {
 			this.refer is s
 		}
 	}
 }

In this file, we have two schemas that model our domain. First, schema Service with an attribute running, procedure start that changes running's value from false to true, and procedure stop that changes running's value from true to false.

We also have schema Client with an attribute refer, which is a reference to an instance of Service. There is a procedure redirect that changes the value of refer with any instance if Service.

Create file task.sfp to hold the task:
```
 include "types.sfp"
 
 initial state {
 	a isa Service {
 		running is true
 	}

 	b isa Service // with "running" is false

 	pc isa Client {
 		refer is a
 	}
 }

 goal constraint {
 	pc.refer is b
 	a.running is false
 }

 global constraint {
 	pc.refer.running is true
 }
```
In this file, we specify a task where in the initial state of our domain, we have two services a and b, and a client pc. a is running, b is stopped, and pc is referring to a. We want to generate a workflow that achieves goal: pc is referring to b and a is stopped, and preserves global constraint: pc is always referring to a running service.

To generate the workflow, we invoke sfp command with argument the path of the task file:

  $ sfplanner task.sfp

Which will generate a workflow in JSON

 {
   "type": "sequential",
   "workflow": [
     {
       "name": "$.b.start",
       "parameters": {
       },
       "condition": {
         "$.b.running": false
       },
       "effect": {
         "$.b.running": true
       }
     },
     {
       "name": "$.pc.redirect",
       "parameters": {
         "$.s": "$.b"
       },
       "condition": {
       },
       "effect": {
         "$.pc.refer": "$.b"
       }
     },
     {
       "name": "$.a.stop",
       "parameters": {
       },
       "condition": {
         "$.a.running": true
       },
       "effect": {
         "$.a.running": false
       }
     }
   ],
   "version": "1",
   "total": 3
 }

This workflow is sequential that has 3 procedures. If you executes the workflow in given order, it will achieves the goal state as well as perserves the global constraints during the execution.

Planner Options

You could set particular environment variable to change the planner settings:

To activate debug-mode

$ export SFPLANNER_DEBUG=1

This will disable automated deletion of temporary files in temporary directory (/tmp/nuri_****/).
To use multiple heuristic on finding the solution, and then pick the best result

$ export SFPLANNER_MIXED_CONTINUE=1
To set heuristics which are used in searching

$ export SFPLANNER_MIXED_HEURISTICS=ff2,cea2

sfplanner

Development

Runtime