Upload
vankhanh
View
220
Download
4
Embed Size (px)
Citation preview
ApproachDrawing on hundreds of hours of in-context observation and experience developing tools1 for the Mars Exploration Rover (MER) mission in 2003, the Scheduling and Planning Interface for Exploration (SPIFe) team began development in 2003 for the Mars robotic missions Phoenix and Mars Science Laboratory (MSL)2. A Bed Rest adaptation of SPIFe was implemented in 2007 with additional features to help the task of scheduling in a human domain. Although planning scheduling for robotic missions seems very different than scheduling in the human domain, it has many of the same core problems. We outline some our lessons learned in the Bed Rest domain here.
Background
The Bed Rest Project
Future Constellation mission concepts will require a new suite of scheduling and planning tools that deliver improvements to current technology and processes, including:
The ability to plan and reason about complex constraint networks on • long-range, mid-range, and tactical real-time schedulesThe ability for human operators to effectively • manipulate and explain schedules generated by automated planning softwareThe ability for • the crew to assume the scheduling and planning roles currently reserved for ground personnel
The Flight Analogs Bed Rest Project simulates the effects of microgravity by having volunteer subjects lie in bed with a head-down 6 degree tilt for 90 days as part of a 120-day campaign.
Planning and Scheduling Software for Bed Rest OperationsLessons Learned Toward a Crew Scheduling System for ExplorationMichael McCurdy1, Jack Li2, Melissa Ludowise1
1NASA Ames Research Center, MS 262-4, Moffett Field, CA 94035 / 2San Jose State University Foundation, NASA Ames Research Center, MS 262-4, Moffett Field, CA 94035
Different Views for Different Tasks
SPIFe has two different ways of presenting the plan based on the user’s task: the Timeline, and the Merge Editor (a tabular view). The Timeline is what the Bed Rest scheduler primarily uses to edit the plan and displays activities in a chronologically in the horizontal direction and by subject in the vertical direction.
Changing the time of an activity done by dragging the activity, represented by a •rectangle, across the screen with the mouse.Information about the activity and timepoint of the mouse are displayed in the •upper left corner.
Timeline Merge Editor
The Importance of Constraints
There are numerous constraints on every activity in a Bed Rest schedule including:Operational Constraints•
Room and equipment restrictions•Traveltimethatittakestophysicallymovethesubjectfromfloortoanotherforatest•
ScientificandHealthConstraints•Drug or meal interactions that would interfere with data•Drug or meal interactions that could make a subject feel ill•
Personal Requests•Lab requests (eg. vacations, preference to not work on a weekend)•Subject requests (eg. eating meals at a certain time of the day)•
In order to ensure all constraints are enforced in a schedule:The scheduler and all other stakeholders refer to a • written list of all constraints when creating and verifying the scheduleValidation that all constraints are being respected was accomplished only by • human checking
Althoughmanypeoplereviewedtheschedulerepeatedly,thefinalschedulestillsometimescontaineduncaught errors which resulted anywhere from a subject getting nauseous to losing or missing the collection of data
Some statistics to illustrate how complex this task is:There are on average between • 5 and 10 subjects in bed at a given time, each of whom is in a different stage of a campaignEach subject has approximately• 10 activities on any given dayThere are currently • 70 different Bed Rest activitiesdefinedandatotalof222 constraints and 21 operational notesdefinedamongthemThiscaneasilyresultinover100conflictsononedayinaschedule•
Toassisttheschedulerwiththetaskofenforcingallconstraintsintheschedule,SPIFenotifiestheuserofallexistingviolationsintheplaninthePlan Advisor
Plan Advisor
Automated Assistance Must Allow for Human Intervention
There are a number of commercial automated planners and schedulers that are designed to resolve any constraint violations in a plan or schedule, usually by moving or removing activities. Most automated planners lack the ability to give the user enough control over what is being changed in the plan and will:
Sometimes take hours or days to solving all the violations in a schedule•Often return an unreasonable schedule without any insight for its reasoning•
SPIFe uses a hybrid approach of automated planning and user control by allowing the user to select aspecificsetofactivitiesintheplantoresolveandbyallowingthemtoacceptorrejectitsproposedchanges for each activity. The Plan Advisor will also always display all the violations that exist in a schedule and allowing the user to resolve them manually without the use of the automated planner.
Fixing Violations in SPIFe
Soft Constraints
In the robotic domain, there exist constraints called flight rules that if broken can result in harm to the vehicle. Alternatively, the Bed Rest schedule had a notion of soft constraints where it is sometimes necessary to “bend the rules”. For example:
A lab may not wish to conduct any tests on a weekend, but if it’s imperative •thatscientificdatabecollectedonaSundaythelaboperatorscanmakeanexception.A subject should be fed lunch between 11:30 AM and 1:00 PM, but it would not •be harmful to allow them to eat slightly earlier or later.
To accommodate for this in SPIFe, the ability for the user to waive flight rules for an activity was added, which would tell the Plan Advisor to ignore the constraints for that activity.
Plan Advisor ignoring waived activity
Unlike the Mars robotic domain where activities are determined months before the mission, the Bed Rest project had a need to add or change activities in the middle of a campaign. For example:
Shortly after beginning the study, it •was discovered that subjects required massage therapy every other day in order prevent their muscles from stiffening.
To facilitate the ability for the Bed Rest team to make changes to activity definitions,anActivity Dictionary Editor is currently under development that would allow the Bed Rest scheduler to edit attributes and constraints and almost immediately use them in SPIFe.
Refinement and Iteration
Activity Dictionary Editor
The ScheduleThe schedule represents the single agreed communication contract of how to operationally execute all the needs of the stakeholders in the Bed Rest project. Paper copies of the schedule are posted in all labs and rooms of the Bed Rest facility so that all nurses, subjects, and lab technicians are able to follow it without risk of violating any constraints.
It’s the responsibility of a single scheduler to plan out the activities for each subject’s daily schedule and take into account:
The • tests that are required on each Bed Rest day in the campaignThe • daily activitieseachsubjectmustperforminadditiontoscientifictests(eg.Meals,Hygiene,Vitals)Any • operational or scientific constraints involving the activitiesPersonal requests•
The scheduler will plan out activities two weeks in advance, send out the schedule on a weekly basis for approval from various stakeholders, and receive feedback for possible change requests. In addition, the scheduler must also change the schedule on short notice if an anomaly occurs when:
Equipment malfunctions• A subject becomes ill• (and cannot perform certain tests)
In the event of an anomaly, the scheduler must alter the schedule, resolveanyconstraintconflictsandsendthescheduleoutbeforethe next morning.
Bed Rest schedule report from SPIFe
Bed Rest schedule from previous operations
Integrated
Detail Schedule
Midnite to 400 Notes 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 2100 Changes vs DCO
15 30 45 15 30 45 15 30 45 15 30 45 15 30 45 15 30 45 15 30 45 15 30 45 15 30 45 15 30 45 15 30 45 15 30 45 15 30 45 15 30 45 15 30 45 15 30 45 15 30 45 15 30 45
1 hr 2 hr 3 hr 4 hr 5 hr 6 hr 7 hr 8 hr 9 hr 10 hr 11 hr 12 hr 13 hr 14 hr 15 hr
Rm. # 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 2100
Sub
7961
BR Day
+0
6.214
No Meds
Sub
3985
BR Day
87
6.238
No Meds No
Acetaminophen Lactulose BR86 to BR87
Sub
7984
BR Day
78
6.212
Sub
6295
BR Day
47
6.230
Sub
8779
BR Day
2
6.238
No Meds No
Acetaminophen
Sub
6112
BR Day
66
6.212
Sub
9274
BR Day
56
6.230
QUS BR56 to BR51
Sub
8250
BR Day
48
6.232
No Meds
Sub
9183
BR Day
-6
6.214
No Meds
Monday, April 30, 2007
Personal
Hygiene
30 min
Breakfast
30 min
Dinner
30 min
Med
Stat
Vitals,
Level 4
No Med
15 min
P Str
etc
hin
g1
0 m
in
Str
etc
hin
g1
0 m
in
Lunch
30 min
Personal
Hygiene
30 min
Functional Stretch Reflex
First FloorLevel 3
60 min
Travel
to 1st
Floor
15 min
Travel
to 6th
Floor
15 min
Dinner
30 min
Med
Stat
Vitals,
Level 4
No Med
15 min
P
Cycle Ergometer
MR080L 6.220Level 1
60 min
Posture Study
Level 3
60 min
Light
Breakfast
Complex
Carbs
30 min
Re
ma
inin
g M
ea
l
Neuro Assessment
MR042L
Bal ControlLevel 1
40 min
Cardiac
Function
(ECHO)6.242
Level 4
25 min
Plasma Volume
Level 3
45 min
Neuroendocrine & Cardio
Response to Tilt
Level 1
65 min
Cardiac
Function
(ECHO)6.242
Level 4
25 min
Light Lunch
Complex
Carbs
30 min
Re
ma
inin
g M
ea
l
Co
mb
ine
d B
loo
d
Dra
w
15
min
L
ev4
Sa
liva
Co
llectio
n
Sb
lng
l S
cra
pe
1
0m
in
Morning Activities:
Saliva Collection
Sublingual Scrape
Combined Blood Draws for
Immunology, Nutrition, and Clinical
Lunch
30 min
Dinner
30 min
Str
etc
hin
g1
0 m
in
Sa
liva
Co
llectio
n
10
min
Personal
Hygiene
30 min
Breakfast
30 min
Lunch
30 min
Dinner
30 min
Med
Status,
Vitals
Level 4
15 min
Str
etc
hin
g1
0 m
in
Str
etc
hin
g1
0 m
in
Personal
Hygiene
30 min
Breakfast
30 min
Lunch
30 min
Dinner
30 min
Med
Status,
Vitals
Level 4
15 min
Str
etc
hin
g1
0 m
in
Str
etc
hin
g1
0 m
in
Sa
liva
Co
llectio
n
10
min
Morning Activities:
Saliva Collection
Personal
Hygiene
30 min
Breakfast
30 min
Lunch
30 min
Dinner
30 min
Med
Status,
Vitals
Level 4
15 min
Vibration
Level 3
30 min
Str
etc
hin
g1
0 m
in
Str
etc
hin
g1
0 m
in
Sa
liva
Co
llectio
n
10
min
Morning Activities:
Saliva Collection
Personal
Hygiene
30 min
Breakfast
30 min
Lunch
30 min
Dinner
30 min
Med
Status,
Vitals
Level 4
15 min
Vibration
Level 3
30 min
MR085L
WINSCATLevel 4
30 min
Str
etc
hin
g1
0 m
in
Str
etc
hin
g1
0 m
in
Sa
liva
Co
llectio
n
10
min
Morning Activities:
Saliva Collection
Personal
Hygiene
30 min
Breakfast
30 min
Lunch
30 min
Dinner
30 min
Vibration
Level 3
30 min
Med
Stat
Vitals,
Level 4
No
Med
15 min Str
etc
hin
g1
0 m
in
Str
etc
hin
g1
0 m
in
Morning Activities:
Saliva Collection
Physiotherapy
60 min
Personal
Hygiene
30 min
Lactulose/Acetaminophen Test
Lact/Acet Breakfast 1st 15 min
Level 3
240 min
La
ctu
lose
/Ace
t L
eve
l 3
15
min
La
ctu
lose
/Ace
t L
eve
l 3
15
min
Re
ma
inin
g M
ea
l
Str
etc
hin
g1
0 m
in
Med
Stat
Vitals,
Level 4
No Med
15 min
P
Personal
Hygiene
30 min
Breakfast
30 min
Dinner
30 min
Lunch
30 min
Isokinetic
First Floor
MR079LLevel 3
60 min
Travel
to 1st
Floor
15 min
Travel
to 6th
Floor
15 min
Med
Stat
Vitals,
Level 4
No Med
15 min
P Sa
liva
Co
llectio
n
10
min
Morning Activities:
Saliva Collection
Physiotherapy
60 min
Physiotherapy
60 min
Physiotherapy
60 min
Physiotherapy
60 min
Travel to
1st Floor
15 min SRC
45 min
Travel to
6th Floor
15 min
BRPfnlApr30-May6rv1.xls9:57 AM
2/1/08
1 Bresina J., Jónsson A., Morris P., and Rajan K. 2005a. “Activity Planning for the Mars Exploration Rovers”, Fourteenth International Conference on Automated Planning and Scheduling, Monterey, 2005, pp. 40-49.2 Aghevli, A.; Bachmann, A.; Bresina, J.; Greene, K.; Kanefsky, B.; Kurien, J.; McCurdy, M.; Morris, P.; Pyrzak, G.; Ratterman, C.; Vera, A.; and Wragg, S. 2006. Planning Applications for Three Mars Missions with Ensemble. In International Workshop on Planning and Scheduling for Space.