An intelligent System for Generation of Automatic Flight Operation Plans for the Satellite Control A

1
An intelligent System for Generation of Automatic
Flight Operation Plans for the Satellite Control
Activities at INPE

• Luciana S. Cardoso
• Maurício G. V. Ferreira
• Valcir Orlando

Rome, 19-23 June 2006
2
Summary

• Introduction
• Planning Agent
• INPE Satellite Operation
• PlanIPOV System
• Architecture
• Prototype Implementation
• Tests and Results
• Limitations Found and Solution Adopted
• Conclusion

3
Introduction

• Research of automatic for space operations is a
  real need for all space agencies in order to
  increase the reliability and reduce cost in space
  operations.
• The planning of satellite operations -gt strategic
  and complex task (minimizes the use of on-board
  resources and maximizes the products of the
  mission without compromising the security of the
  satellite).
• Planning Technology from the AI -gt a solution for
  automatic plan generation.

4
Motivation

• A significant parcel of satellite operation
  activities at INPE is still performed manually -gt
  finding automated alternatives for the operation
  is of capital importance.

(Increased in the number of satellites to be
controlled Scarcity of financial resources
Maintenance of the currently satisfactory
performance of the operation activities.)
PlanIPOV Intelligent planning system for
automatic generation of satellite flight
operation plans for the routine operation phase
based on planning technology and using PDDL 2.2
language.
5
Planning Agent
6
Architecture of Planning Agent using PDDL
7
INPE Satellite Operation
8
INPE Satellite Operation
9
Architecture of PlanIPOV System
10
Prototype Implementation
11
Prototype Graphic User Interface
12
Tests and Results

• Results of comparison between the prototype FOP
  files and the real satellite FOP files.

13
Limitation Found x Solution Adopted

• 1st Limitation
• Generation of problem files at each change of
  environment (every pass) gt very high cost in
  terms of time and effort consumption.
• Solution
• Automatic generation of the problem file for each
  pass to be tracked -gt Configuration and Problem
  Generator modules.

14
Limitation Found x Solution Adopted

• 2nd Limitation
• The structure of the PDDL did not allow the
  classification of predicates in Domain files -gt a
  generic solution could not be programmed to
  automates generation of the Problem files.
• Solution
• Definition of predicate categories (initial
  state, internal events, exogenous events and
  goals states) and insertion of the name of each
  category in the Domain file as a commentary line.

15
Predicate Categories in the Domain File

• (predicates
• Internal Events
• (TCSent ?pass - PassNumber ?sat - SatCode)
• (CalibrationOK ?pass - PassNumber ?cgroup -
  CalibGroupNumber)
• Initial Environment States
• (GSAvailable ?gs - GSCode)
• (PassTrack ?pass - PassNumber ?sat - SatCode ?gs
  - GSCode)
• Exogenous Events
• (CalibrationAvailable ?pass - PassNumber ?gs -
  GSCode)
• (FirstTc ?pass - PassNumber ?sat - SatCode ?tc -
  TCCode)
• (LastTC ?pass - PassNumber ?sat - SatCode ?tc -
  TCCode)
• Goals
• (FirstRangeOk ?pass - PassNumber ?rgroup -
  RangeGroupNumber)
• (PassEliminated ?pass - PassNumber ?sat -
  SatCode ?gs - GSCode)
• (LastTCOK ?pass - PassNumber ?sat - SatCode ?tc
  - TCCode)
• )

16
Limitation Found x Solution Adopted

• 3rd Limitation
• The PDDL structure does not allow defining
  priorities for the exogenous events in the
  problem files.
• Solution
• A C code was implemented to establish a
  priorities among the exogenous events.

Problem without solution.
17
Limitation Found x Solution Adopted

• 4th Limitation
• The PDDL structure does not allow the
  establishment of priorities for the sub-goals to
  be reached in the plan.
• Solution
• A C code was implemented to discard the
  sub-goals that exceeded the period of time
  comprised by a pass.

18
Conclusion

• This work is considered as a first step in
  direction to get higher automation level for the
  execution of satellite control activities.
• The use of the technology of planning agents is a
  promising way to reduce costs of the satellite
  operation at INPE and increase the reliability.
• The definition of the PlanIPOV System
  architecture and the implementation of the
  prototype allowed generating FOPs for all
  satellites actually controlled by INPE.
• Great advantage -gt it was possible to comply with
  the operation characteristics of all satellites
  only by modifying the configuration database. No
  changes in the code.
• This feature gives flexibility on the FOPs
  generation during the operators training phase -gt
  to allow the variation in the initial environment
  and in the goals.

19
Next Step

• Executor Module
• Meaningful reduction of satellite operation
  efforts.

20
Luciana S. Cardoso Mauricio G.V.
Ferreira luciana_at_dss.inpe.br mauricio_at_ccs.inp
e.br Valcir Orlando valcir_at_ccs.inpe.br