From: AAAI Technical Report SS-95-04. Compilation copyright © 1995, AAAI (www.aaai.org). All rights reserved.
Planning for Deep Space Network Operations
Randall Hill, Jr., Steve Chien, Crista Smyth, Kristina Fayyad, Trish Santos
Jet PropulsionLaboratory
CaliforniaInstitute of Technology
4800 Oak GroveDrive, M/S525-3630
Pasadena, CA91109-8099
{ hill, kristina, crista, Irish }@negev.jpl.nasa.gov
chien@aig.jpl.nasa.gov
Abstract
Weare developing a plan construction and plan
executionmonitoringsystemto automatethe operation
of the communicationslink in NASA’sDeep Space
Network(DSN).Theobjectives of this effort are: (1)
to archive knowledge
of operationsprocedures;(2)
providea capability to assist in executionmonitoring;
and (3) provide a capability for automatically
constructing the plan to be executed and monitored
based on an augmentedsequence of events (SOE).
This paper describes the application area of DSN
antennaoperations, describesthe plan representation
of a TemporalDependenceNetwork(TDN),outlines
the knowledgeengineeringtools being constructedto
facilitate in representingDSNoperationsprocedures,
and describes preliminary work in automated TDN
generation.
1. Introduction
Oneof the significant challenges to developingand
fielding a real-worldplanneris in the representatiofi and
acquisition of the knowledgeneededfor it to function
autonomouslyin a real-world problemdomain.This paper
describes howweare addressingthe knowledge
acquisition
and representation problem in the developmentof an
intelligent system for representation of procedural
knowledgerepresentation, plan execution monitoring,and
plan construction, for a real-world domain: the
communications link monitor and control system of
NASA’sDeep Space Network (DSN). The underlying
motivation driving this automation is improving the
productivity of the DSNwhile maintaining at least the
current level of reliability. To accomplish this goal
automatedplan generation and execution monitoringmust
be capable of performingmanyof the sameoperations that
are currently donemanuallyby DSNstation Operators, and
thereby significantly reduce the amountof humaneffort
neededto performmonitorandcontrol tasks in the domain.
Wehaveelsewhereidentified a numberof factors that
makeit difficult to implement
a plannerfor a domainlike
the DSN(see Chien, Hill, andFayyad,1994). In this paper
wedescribe our efforts to developa plan construction,
executionand monitoringsystemto automatethe operation
of the communications links in NASA’sDeep Space
Network(DSN).Theobjectives of this effort are: (1)
archive knowledge
of operationsprocedures;(2) to provide
a capability to assist in executionmonitoring; and (3)
providea capability for automaticallyconstructingthe plan
to be executed and monitored based on an augmented
sequenceof events (SOE).
Webegin by giving a brief overviewof the DSNtask
domainand the nature of problemsolving there. Next, we
describe howweare automatingthe operation of the DSN
throughthe use of a plan executionand monitoringsystem
called the Link Monitorand Control Operator Assistant
(LMCOA),
whichuses a plan representation schemecalled
a TemporalDependencyNetwork(TDN).A description
given of howplan knowledgewill be captured using some
knowledgeacquisition tools that are currently under
development.
Next, in the section entitled "SOE-driven
automation"
wedescribe the plannerthat will eventuallybe capableof
generatingall of the plans neededto operate the DSN
for a
widevariety of telecommunications
services. Wecall this
"SOE-drivenautomation"becausethe input to the planner
is a sequenceof events (SOE)that describesthe basic track
being requested. Theplanner uses the SOEto create goals
that must be satisfied by the plan it generates. Examples
from the DSNdomainare given that showthe kinds of
information the planner workswith as well as what it
produces.
2. Deep Space Network Domain
The Deep Space Network (DSN) is a worldwide
network of. deep space tracking and communications
complexeslocated in Madrid,Spain, Canberra,Australia,
and Goldstone, California. Each of these complexesis
capable of performingmultiple missions simultaneously,
each of whichinvolves operating a communications
link. A
DSNcommunications
link is a collection of devicesused to
track and communicate
with an unpilotedspacecraft.
Currently,mostof the tasks requiring the control of a
DSNcommunications link are performed by human
Operators on a system called the LMC
(Link Monitorand
Control)system. TheOperatorsare given tasks that involve
configuring and calibrating the communications
equipment,
and then they monitor and control these devices while
tracking a spacecraft or celestial body. The Operators
followwritten proceduresto performtheir missiontasks. A
procedurespecifies a sequenceof actions to execute, where
the actions are usually commands
that mustbe entered via
the link’s monitor-and-control
systemkeyboard.
Once issued, a commandis forwarded to another
subsystem,whichmayaccept or reject it, dependingon the
state of the subsystemat the time that the command
is
received. TheOperator receives a messageback from the
subsystemindicating whether the command
was accepted
or rejected, and in cases wherethere is no response, a
messagesaying that the command
"timed out" is sent.
These messagesdo not indicate whether the action was
successfulor whatthe results of the action were. Rather,
the Operator has to monitor subsystem displays for
indicationsthat the actioncompleted
successfullyandthat it
hadits intended effects. It is common
for commands
to be
rejected or for commands
to fail due to a numberof realworldcontingenciesthat arise in the executionof a plan or
procedure.
3. Automatingthe DSN
A prototype called the Link Monitor and Control
Operator Assistant (LMCOA)
was developed to improve
Operatorproductivity by automatingsomeof the functions
of the LMC.The LMCOA
performstasks by: (1) selecting
a set of plans to execute, (2) checkingwhethera plan’s
preconditions have been satisfied, (3) issuing the
commands,and (4) subsequently verifying that the
commandshad their intended effects. The Operator
interacts with the LMCOA
by watchingthe plans as they
are being executedand pausingor skippingportions of the
plan that needto be modifiedfor somereason. Whena plan
fails, the LMCOA
lacks the ability to recover on its own.
Instead, the Operatoris left to figure out howto recover
from the failure. In observing Operators performtheir
duties, wefind that they havean amazingability to adapt
the written proceduresto the contingenciesthat arise in the
courseof their execution.Areal-worldplannerhas to have
the sameflexibility and knowledge
as a humanOperatorin
order to succeedin performingDSNmissions.
Temporal Dependency Network (TDN)
Weuse a representation called a TemporalDependency
Network(TDN)to express the basic plan structure and
interrelationships amongplans. ATDN
is a directed graph
that incorporates temporaland behavioral knowledgeand
also provides optional and conditional paths throughthe
network.Thedirected graphrepresents the steps required
to performan operation. Precedencerelations are specified
by the nodes and ares of the network. The behavioral
knowledge
identifies system-statedependenciesin the form
of pre- and postconditions. Temporalknowledgeconsists
of both absolute (e.g. Acquire the spacecraft at time
02:30:45)andrelative (e.g. Performstep Y5 minutesafter
step X) temporalconstraints. Conditionalbranchesin the
network are performedonly under certain conditions.
Optional paths are those whichare not essential to the
operation, but may,for example,providea higher level of
confidencein the data resulting froma plan if performed.
Eachnodein the TDN
is called a plan and contains actions
to be performed. A plan also has goals, pre- and
posteondition constraints and temporal constraints
associatedwith it. Moredetails about TDNs
are providedin
(Fayyad& Cooper1992).
52
In our experience of building TDNs,the knowledge
bases associatedwith evena single TDN
are quite large and
difficult to build andmaintain. In addition, they rely on
informationfrom manydifferent sources: documentation
in
paper form, interviews with experts, and text logs of
operations proceduresfromactualmissions, just to namea
few. As we build more TDNs,muchof this information
can be usedagain.
KnowledgeAcquisition Tools
Several tools are under developmentto assist in the
acquisition and maintenanceof the plan knowledgebase
(for moredetails see Hill et al., 1994). ATDN
authoring
tool is being developedto automatethe specification of
TDNs.Developersas well as operations personnel will be
able to graphically specify the TDN
and its contents. TDNs
can be composed
from parts of existing TDNsand libraries
of actions at the least. A databasewill efficiently store a
completespecification of a TDN
as part of a TDN
library.
Thesamedatabasewill serve as a central repository for the
TDNin the LMCOA,thus simplifying the LMCOA
implementation.In addition, the TDNauthoring tool will
includethe capability to verify certain aspects of the TDN
such as incompatible block ordering based on pre and
postcondition constraints of blocks. This TDNdatabase
will also allowTDN
developersto quicklyand easily access
TDN
informationrelated to the TDNs,blocks, subsystems,
and directives being modified. For example, when
constructing a newTDN
block, the developerwill be able
to easily accessother blockscontainingthe samedirectives
as well as accessother blocksaffecting the samesubsystem
state variables as the current block. The knowledge
engineeringeffort for the LMCOA
prototypeis describedin
moredetail in (Fayyadet al., 1993).
Withthe TDNauthoring tool, users can graphically
build a TDN
and specify all the informationrequired in a
TDN. For example, they graphically create boxes
representingplans, and drawthe links betweenthe plans to
represent dependenciesbetweenplans. A small segmentof
a TDNis shownin Figure 2 and contains three plans in
sequence.Thismaybe all that the users sees if she initially
creates just a high level TDN
includingskeletal plans anda
partial order of plans. Whena plan is created, the user
gives it a nameandspecifies whattype of plan it is; there
are different types of plans depending
on the type of actions
to be takenin a plan, e.g. a manualactionto be takenby the
Operator,or a list of commands
to be sent to a subsystem.
In our example, each of these plans is a command
plan
associated with the 34 meter Antenna Subsystem. The
sequenceof actions for the plan namedClear Boresight
Offsets is shownin Table1. Thefirst twoactions set the
Elevationand CrossElevationcorrection offsets to 0. The
last action sends the structure containing these two
variablesto the subsystem.
howto control them under both normal and anomalous
conditions.
4. SOE-Driven Automation
Plan: Clear Boresight Offsets
Actions: setvar A34CorrEl0
setvar A34CorrXel0
sendvar ActualCorrData
Table1. Planactions.
I
Clear BoresightOffsets
sendvar ActualCorrData
ActualCorrData.el= 0
ActualCorrDatax.el= 0
Table2. Postconditionfor Clear BoresightOffsets.
Plan:
Action:
Postcondition:
For eachplan, the user can then specifythe sequenceof
actions withinthe plan, the pre- andpostconditionsof each
action, goals of plans, andtemporalconstraints on plans.
Dependingon the knowledgethat the user has about the
actions, he mayor maynot be able to specify all of this
information.Thepostconditionsof the last action in Table
1. fall out nicely fromthe action itself; thoughthis isn’t
alwaysthe case. Since this last action sends data to the
subsystem,it is necessaryto makesure that the subsystem
did indeedreceivethe newdata values. In this case, weare
only concernedwith the two variables in the structure
whichcorrespondto the f’nost twoactions in the plan. The
postconditions for the last command
in Clear Boresight
Offsetsare shownin Table2.
In our current prototype,mostof the informationabout
a plan is specified by filling in text forms. As the tool
develops,knowledge
bases of actions with their associated
pre- and postconditionsand knowledge
bases of plans will
be available so that newTDNs
can be built frompieces of
existingones.
Theoutputs of the tool are a text file that completely
defines the TDNand a file depicting the graphical
representation of the TDN.Thesetwo files will be loaded
by the executionmonitor. Thegraphical representation is
the basis for a dynamicdisplay of the TDN
in the execution
monitor which shows the progress of the TDNas it is
executing.Thetext file is loadedby the executionmonitor
whichuses it to create an executableTDN.
Besides the TDNauthoring tool, two other tools,
RIDESand REBUS,are being developed and used for
knowledgeacquisition. The RIDESsimulation authoring
tool kit (Munroet al., 1993) is used to capture device
models of the communications link equipment and
subsystems.Besidesusing these modelsin the planner, the
simulator also permits us to test the LMCOA’s
ability to
copewith the operational context issues described in the
previous section. REBUS,
whichstands for Requirements
EnvisioningByUtilizing Scenarios, (Zorman,1994)is used
to capture knowledgeabout the domainby using different
scenarios to provide contextual information needed for
planning. This provides us a wayto understand howthe
subsystems controlled by the LMCOA
actually work and
53
While the currently operational LMCOA
uses handconstructed, manuallyselected plans froma plan library,
the current work focuses on automated generation of
operations procedures for DSNantenna operation. This
capability is to be demonstrated
at the endof January1995.
In this task, severalsourcesof information
are available:
- Project SOE- The project sequence of events
specifies eventsfromthe mission/projectperspective. Asa
result, the project SOEcontains a great deal of information
regardingthe spacecraft state whichis relevant to the DSN
track, as well as a large amountof spacecraft information
unrelated to DSNoperations. Relevant information
specifiedin the project SOEincludessuch itemsas the onewaylight time (OWLT)
to the spacecraft, notifications
the beginningand endingtimes of tracks, spacecraft data
transmission bit rate changes,and carrier and subcarrier
frequency changes. An excerpt from a VOYAGER
project
SOEis shownin Figure4.1.
- Projectprofile - This file specifies project specific
information regarding frequencies and pass types. For
example,the Project SOEmightspecify bit-rate = HIGH,
andthe project profile wouldspecifythe exactbit rate used.
Theproject profile mightalso specify exact frequencies
used,anddefault track types.
- LSOE- Usingthe project prof’de information, the
project SOEis parsed into a more manageable form
specifying only the informationdirectly relevant to the
specific pass. A sample LSOEfor a VOYAGER
telemetry
downlinkpass is shownin Figure4.2 below.
- TDN KB - The TDNknowledge base stores
information on the TDNblocks available for the DSN
Planner and LMCOA
to use. This database includes
information regarding preconditions, postconditions,
directives, and other aspectsof the TDN
blocks.
* ID=-DSNISOE52B1 S/C--032 BEG=94360 094000
END=95001 221500
1 032 42 00022 360 201500 ACENCTS/C ANT
STATUS,RCP/00.100/Y, HGA,RCP/00.100/Y, HGA
1 032 42 00022360 201500ACENCT
S/C TXR
STATUS,OFF, EXC1/TWT2/18, LOW,EXC1/TW’I2/18
1 032 42 00022 360 201500ACENCT S/C RCVR
STATUS,2/S, ON
1 032 42 00022 360 201500ACENCT S/C RNG
MODE,OFF, OFF, ON, OFF
1 032 42 00022 360 201500ACENCT S/C TLMXMOD,160CD, LOW,61DG
1 032 42 00022360 201500 I)42 BOT
1 032 42 00022360 201500 D42 ACQD/L, 1-WAY,X
! 032 42 00031361 095000 1342 EOT
Figure 4.1: Sample VOYAGER
Project SOE
planning techniques to produce a parameterized track
95/005/16:30:00 ACQ_D/L:MODE=l-WAY;
BAND=X;.
specific TDNto be used to conduct the track. The actual
95/005/16:30:00 CONFIG:
Ib’3=S; IF4=X; POLAR=RCP;
SKY_FREQ=10000000;PRED_POWER=50;
planner used to generate the TDNis a modified version of
PO_FILE="/u/cm/lmcoa/lmcoa/src/kb/sourco.po";.
the task reduction planning componentof the MVPsystem
95/005/16:30:00 TPR_PRECAL:
BAND=S,X;.
(Chien 1994). This track-specific TDNand the LSOEcan
95/005/17:15:00 SOURCE:
SRC=VGR2;.
then be used by the LMCOA
system to operate the actual
antennas to conduct the requested antenna mack.
95/005/17:15:00 CARRIER:
PC_NO=34;
FREQ=70000000;
For example, consider the task reductions shown in
BANDWlDTH=I.0;
UPDATE_RATE=500;
LOOP=-2;.
95/005/17:15:00 SUBCARRIER:
PD_NO=39;FREQ=-21333;
Figure 4.5 and 4.6. As part of the pre-calibration procedure
BANDWIDTH--1.0;
UPDATE_RATE=100;
LOOP=-2;.
for the station DSS13
for a downlinktype of track, there are
95/005/17:15:00 SYMBOL:
ES_.NO=40;SYMBOL_RATE=500; a series of steps whichmust be taken. As Figure 4.5 shows,
BANDWlDTH=I.0;
UPDATE_RATE=100;
LOOP--2;
these tasks are inspecting the subsystems, connecting the
CODE_FLAG=l;.
subsystems, configuring the total powerradiometer, loading
95/005/17:15:00 TELEM:B1T_RATE=10;
MOD_INDEX=20;. antenna predicts, and configuring the receiver. Someof
95/005/18:00:00 TELEM:MOD_INDEX=30;.
these tasks are compositetasks, and will be later expanded
into moredetailed tasks. For example, configuring the total
95/005/18:30:00 TELEM:
BIT_RATE=20;.
95/005/19:00:00 TELEM:BIT_RATE=50;MOD_INDEX=60;. power radiometer involves configuring the IF switch,
configuring the UWC/TPRfor pre-calibration,
and
95/005/20:00:00 END_OF_TRACK.
performingthe actual TPRpre-calibration. In the context of
this specific track, the programmableoscillator (PO) has
Figure 4.2: VOYAGER
downlink telemetry pass LSOE
beenrequested, so that additional steps will be inserted into
the TDNas shownin Figure 4.6. In the context of specific
The overall flow of information in our approach to
tracks, generic configuration blocks will be specialized as
SOEdriven automation is shown in Figure 4.3. In this
appropriate to the details of the track. For example,in the
approach, the project SOEis parsed into an LMCOA
SOE
(LSOE)which contains only that information relevant
context of the VOYAGER
track outlined above, the track
type is downlink, telemetry, with symbol decoding rethe track of interest. The DSNplanner uses the LSOEand
quested. This will require a specific type of receiver conthe Project profile to determine the parameterized goals of
figuration block. This process continues until all of the acthe track. A sample set of track goals is shownbelow in
tivity blocks in the TDNare operational - that is to say that
Figure 4.4. This set of goals and a set of contextual facts
derived from the LSOEare the planning inputs. The
knownblocks in the TDNKBcan be used to instantiate the
activities. This fully instantiated TDNcan then be used
contextual facts are basically the planning system’s
with the LSOEby LMCOA
to perform the actual track.
representation of the LSOEinformation.
ProjectSeE
Station DSSt3 [ pre-cagtbretlon]
Project Profile
downllnktrack
)
TON KB
Figure 4.5: Sample Task Reduction
DSN Planner
Station DSS13
LMCOA
POrequired
¢
TrackSpecificTDN
connect
_ I load
I _ I c°nfigurol
_ load
subsystams~l~ 1 POfiles ~ doppler ~ antenna
| tuner
| predicts
Figure 4.6: Augmentation of TDNRequired by Additional
ProgrammableOscillator Track Goal
DS$
Figure 4.3: SOEDriven AutomationArchitecture
(PO_required trackl)
(spacecrafi._mode_changcstrackl)
(track_goal spacecraft_track telemetry trackl)
(wack._goalspacecraft_track downlinktrackl)
(track..goal
decode_data)
(station-used
trackl
DSS13)
Figure 4.4: ExampleTracking
Goals
track
downllnk
~~~,~
I configure
symbols
telemetry
configure
I receiver
Figure 4.7: A Specialized Task Reduction
Using this problem specification, the DSNplanner
then uses task reduction planner (also called
hierarchical
task network or HTN)(Erol et al., 1994; Lansky, 1993)
5~
decoding
Figure 4.8: TDNfor a VOYAGER
telemetry downlinktrack using the 34-meterbeamwaveantenna
Considerable effort in computingthe final TDNis
devotedto determiningthe correct parametersfor blocksin
the TDN.For example,Figure 4.9 showsa configuration
table usedto determinethe IF switchparameterfor the TPR
precalibration step. Dependingon the communications
bands used in the track, differing bandswill assignedto
each of the communicationspathways in the UWC/TPR.
Based on the bands being used, the TPR IF switch
parameteris also determined.This parametersetting is also
determinedduringthe decomposition
process, so a correctly
parameterizedTDNcan be constructed.
IF1
IF2
Parameter
S -band
1
X-band
1
Ka-band
2
2
Q-band
S&X-band
3
X&Ka-band
3
Figure 4.9: TPRIF SwitchParameterDetermination
5, Conclusions
This paper has describedan effort to developa plan
construction and plan execution monitoring system to
automate the operation of the communicationslink in
NASA’sDeep Space Network (DSN). The objectives
this effort are: (1) to archive knowledgeof operations
procedures;(2) to providea capabilityto assist in execution
monitoring;and (3) provide a capability for automatically
constructing the plan to be executedand monitoredbased
on an augmentedspacecraft sequence of events (SOE).
This paper described the application area of DSNantenna
operations, describedthe plan representationof a Temporal
Dependence Network (TDN), outlined the knowledge
engineering tools being constructed to facilitate in
representing DSNoperations procedures, and described
preliminary workin automatedTDN
generation.
6. References
Thus,usingthese decomposition
rules, a track-specific
TDNcan be derived. Figure 4.8 shows a TDNfor a
VOYAGER
downlink telemetry track using the
Programmableoscillator for the DSS13DSNantenna
station. This parameterizedTDNcan then be used by the
LMCOA
system to conduct the actual spacecraft downlink
track.
55
(Chien, Hill, Fayyad,1994) S. Chien, R. Hill, &
Fayyad. "WhyReal-worldPlanning is Difficult." Working
Notes of the 1994 Fall Symposiumon Planning and
Learning: On to Real Applications, NewOrleans, LA,
November1994, AAAIPress.
(Chien, 1994b) S. Chien, "AutomatedSynthesis
ComplexImage Processing Procedures for a Large-scale
ImageDatabase,"Proceedingsof the First IEEEInt. Conf.
on ImageProcessing, Austin, TX,November1994, Vol 3,
pp. 796-800.
(Erol et. al. 1994) K. Erol, J. Hendler, and D. Nau,
"UMCP:
A Soundand CompleteProcedure for Hierarchical
Task Network Planning," Proceedings of the Second
International Conferenceon AI PlanningSystems, Chicago,
IL, June1994,pp. 249-254.
(Fayyad & Cooper, 1992) Fayyad, K. E. and L.
Cooper. "Representing Operations Procedures Using
Temporal Dependency Networks," SpaceOPS ’92,
Pasadena, CA,November1992.
(Fayyadet al., 1993) K. Fayyad,R. Hill, and E.J.
WyatL "KnowledgeEngineering for TemporalDependency
Networks as Operations Procedures." Proceedings of
AIAAComputingin Aerospace9 Conference, October 1921, 1993,San Diego, CA.
(Lansky,1993) A. Lansky,"Localized Planning with
Diverse Plan Construction Methods,"Technical Report
FIA-93-17,NASA
AmesResearch Center, June 1993.
(Hill et al., 1994)Hill, R.W.,Jr., Fayyad,K., Santos,
T., Sturdevant, K. "Knowledge
Acquisition and Reactive
Planningfor the DeepSpaceNetwork."Proceedingsof the
1994 AAAIFall Symposium
on Planning and Learning: On
to Real Applications,NewOrleans, LA,1994.
(Hill&Johnson, 1994) R. Hill and W.L. Johnson.
"Situated Plan Attribution for Intelligent Tutoring."
Proceedings of the Twelfth National Conference on
Artificial Intelligence, July 31-August4, 1994,Seattle,
WA.
(Munroet al., 1993) Munro, A., Johnson, M.C.,
Surmon,D.S., and Wogulis, J.L.. "Attribute-Centered
SimulationAuthoringfor Instruction. "proceedingsof AIED93, WorldConferenceon Artificial Intelligence in
Education,Edinburgh,Scotland; 23-27 August1993.
(Zorman, 1994) Zorman, Lorna. "Requirements
Envisaging by Utilizing Scenarios." Ph.D. Diss., in
preparation,Universityof SouthernCalifornia, 1994.