Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2...
Transcript of Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2...
![Page 1: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/1.jpg)
courtesy of JPL
Programs that Monitor Hidden State: Mode Estimation and Conflict-directed A*
Brian C. Williams16.412J / 6.834J
February 8th, 2016Brian C. Williams, copyright 2000
This image is in the public domain.1
![Page 2: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/2.jpg)
Today’s AssignmentsProblems Sets:• Problem Set #1, Out today, due Wed, February 17th.
Readings:• Today: B. C. Williams, and R. Ragno, "Conflict-directed A* and its Role in
Model-based Embedded Systems," Special Issue on Theory and Applications of Satisfiability Testing, Journal of Discrete Applied Math, January 2003.
• Wednesday: Same.
Background:• 16.410/13 Lectures on Informed Search, Constraint Satisfaction, Propositional
Satisfiability and Diagnosis.
2
2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State
![Page 3: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/3.jpg)
Outline
• Programs that monitor and control hidden states.
• Consistency-based Diagnosis
3
2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State
![Page 4: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/4.jpg)
Base Station
Program sequence of actions in RMPL
method run() {
sequence {
uav.launch();
uav.fly_to_base_station();
uav.pick_up_med_kit();
uav.fly_to_hikers();
uav.drop_off_med_kit();
}
}
Actions have preconditions &
effects like before
Programs that Monitor State
4
2/9/15 16.412J/6.834J S16: L2 Monitoring Hidden State
![Page 5: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/5.jpg)
A Traditional ReactiveProgramming Language
Expressions: 1. s Conditions on sensors2. u Assignments to control variables
Control constructs:1. u
2. If s next A3. Unless s next A4. A, B5. Always A
where A, B are programs.
Action Model: PDDL
Control assignmentsConditional executionPreemptionFull concurrency Iteration
5
![Page 6: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/6.jpg)
senseP(s)
WORLD
observations actions
AGENT
Self-Repairing Agent:
• Monitors & Diagnoses
• Repairs & Avoids
• Probes and Tests
Plant
act
Symptom-directed
Programs that Monitor and Control Hidden (Failure) States
6
10/24/11 copyright Brian Williams, 2000-2012
![Page 7: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/7.jpg)
engine to standby
planetary approach
separate
lander
switch to
inertial nav rotate to entry-orient
& hold attitude
Switch navigation mode:
“Earth-relative” = Star Tracker + IMU
Switch navigation mode:
“Inertial” = IMU only
Mission Storyboards Specify Evolving States
7
![Page 8: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/8.jpg)
engine to standby
Rotate spacecraft:
• command ACS to entry orientation
planetary approach
separate
lander
switch to
inertial nav rotate to entry-orient
& hold attitude
Mission Storyboards Specify Evolving States
8
![Page 9: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/9.jpg)
Like Storyboards, Model-based ProgramsSpecify the Evolution of Abstract States
Embedded programs evolve actions
by interacting with plant sensors
and actuators:
• Read sensors
• Set actuators
Embedded Program
S
Plant
Obs Cntrl
Model-based programs evolve
abstract states through direct
interaction:
• Read abstract state
• Write abstract state
Model-based
Embedded Program
S
Plant
Model-based executive maps
between state and sensors/actuators.
S’Model-based Executive
Obs Cntrl
Programmer maps between state
and sensors/actuators.9
![Page 10: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/10.jpg)
Model-based Programming of a Saturn Orbiter
Turn camera off and
engine on
EngineA EngineB
Science Camera
OrbitInsert()::
do-watching (EngineA = Thrusting OR
EngineB = Thrusting)
parallel {
EngineA = Standby;
EngineB = Standby;
Camera = Off;
do-watching (EngineA = Failed)
{when-donext (EngineA = Standby) AND
Camera = Off)
EngineA = Thrusting};
when-donext (EngineA = Failed AND
EngineB = Standby AND
Camera = Off)
EngineB = Thrusting}
10
10/24/11 copyright Brian Williams, 2000-2012
![Page 11: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/11.jpg)
The program assigns EngineA = Thrusting, and the model-based executive . . . .
Determines that valves
on the backup engine B
will achieve thrust, and
plans needed actions.
Deduces that a valve
failed - stuck closed
Plans actions
to open
six valves
Fuel tankOxidizer tank
Deduces that
thrust is off, and
the engine is healthy
Prog: EngineB = Thrusting
1110/24/11 copyright Brian Williams, 2000-2012
![Page 12: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/12.jpg)
Plant Model:Probabilistic Constraint Automata (PCA)
Standby
Engine Model
Off
Failed
Firing
component modes…
(thrust = full) AND
(power_in = nominal)
(thrust = zero) AND
(power_in = zero)
(thrust = zero) AND
(power_in = nominal)
described by finite domain constraints on variables…
guarded deterministic and probabilistic transitions
off-
cmdstandby-
cmd
0.01
0.01
standby-
cmd
fire-
cmd
cost / reward & prior distribution
0 v
0 v
2 kv
2 kv
one per component … operating concurrently
On
Camera Model
Off
turnoff-
cmdturnon-
cmd
(power_in = zero) AND
(shutter = closed)
(power_in = nominal) AND
(shutter = open)
0 v
20 v
0.01
0.01
0 v
[Williams & Nayak 95, Williams et al. 01]
1210/24/11 copyright Brian Williams, 2000-2012
![Page 13: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/13.jpg)
A Reactive Model-basedProgramming Language (RMPL)
Idea: A concurrent constraint language (e.g. TCC/HCC [Saraswat et al.])• whose constraints c operate on the state of the plant s, and• replaces the constraint store with a model-based controller:
1. c[s]
2. If c[s] next A3. Unless c[s] next A4. A, B5. Always A
Action Model:Probabilistic ConstraintAutomata
Primitive constraint on stateConditional executionPreemptionFull concurrency Iteration
13
![Page 14: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/14.jpg)
Control Sequencer
Deductive Controller
System Model
CommandsObservations
Control Program
Plant
Titan Model-based ExecutiveRMPL Model-based Program
State goalsState estimates
Generates target goal states
conditioned on state estimates
Mode
Estimation
Mode
Reconfiguration
Tracks
likely
plant states
Tracks least
cost goal states
Executes concurrently
Preempts
Queries (hidden) states
Asserts (hidden) state
OrbitInsert()::
(do-watching ((EngineA = Firing) OR
(EngineB = Firing))
(parallel
(EngineA = Standby)
(EngineB = Standby)
(Camera = Off)
(do-watching (EngineA = Failed)
(when-donext ( (EngineA = Standby) AND
(Camera = Off) )
(EngineA = Firing)))
(when-donext ( (EngineA = Failed) AND
(EngineB = Standby) AND
(Camera = Off) )
(EngineB = Firing))))
Closed
Valve
OpenUn-
known
Stuck
closed
Open Close
0. 01
0. 01
0.01
0.01
inflow iff outflow
1410/24/11 copyright Brian Williams, 2000-2012
![Page 15: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/15.jpg)
Deductive Controller
CommandsObservations
Plant
State goalsState estimates
Mode
Estimation:
Tracks likely
States
Mode
Reconfiguration:
Tracks least-cost
state goals
Optimal CSP:
arg min f(x)
s.t. C(x) is satisfiable
D(x) is unsatisfiable
arg min Pt(Y| Obs)
s.t. Ψ(X,Y) O(m’) is consistent
arg max Rt(Y)
s.t. Ψ(X,Y) entails G(X,Y)
s.t. Ψ(X,Y) is consistent
s.t. Y is reachable
Mode Reconfiguration:
Select a least cost set of commandable
component modes that entail the current
goal, and are consistent.
Mode Estimation:
Select a most likely set of next
component modes that are consistent
with the model and past observations.
15
![Page 16: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/16.jpg)
Outline
• Programs that monitor and control hidden states.
• Consistency-based Diagnosis
16
2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State
![Page 17: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/17.jpg)
Estimating Failure Modes Requires Reasoning
from a Model: STS-93Symptoms:
• Engine temp sensor high
• Oxygen level low
• Guidance detects low thrust
• Hydrogen level possibly low
Problem: Liquid hydrogen leak
Effect:
• LH2 used to cool engine
• Engine runs hot
• Consumes more LOX
This image is in the public domain.17
2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State
![Page 18: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/18.jpg)
Model-based DiagnosisInput: Observations of a system with symptomatic behavior,
and a model Φ of the system.
Output: Diagnoses that account for the symptoms.
1 Symptom1
0
A
B
C
D
E
F
G
X
Y
Z
1
1
1
0
1
0
11
1
A1
A2
A3
X1
X2
A1
X1
18
2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State
![Page 19: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/19.jpg)
How Should Diagnoses Account for Novel Symptoms?
Consistency-based Diagnosis: Given symptoms, find diagnoses that are consistent with symptoms.
Suspending Constraints: For novel faults, make no presumption about faulty component behavior.
1
0
1 SymptomA
B
C
D
E
1
1
1
0
1
F
G
X
Y
Z
0
1
A1
A2
A3
X1
X2[Davis, 84]
[Geneserth, 84]
[deKleer & Brown, 83]
19
2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State
![Page 20: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/20.jpg)
Issue 3: Multiple Faults Occur
• three shorts, tank-line and pressure jacket burst, panel flies off.
Diagnosis = Mode AssignmentSolution: Divide & Conquer
courtesy of NASAAPOLLO 13
This image is in the public domain.
20
2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State
![Page 21: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/21.jpg)
Solution: Identify all Combinations of Consistent “Unknown” Modes
• Candidate: Assignment of G or U to each component.
And(i): G(i):
Out(i) = In1(i) AND In2(i) U(i): No Constraint
Candidate = {A1=G, A2=G, A3=G, X1=G, X2=G}
A
B
C
D
E
1
1
1
0
1
F
G
X
Y
Z
0
1
A1
A2
A3
X1
X2
21
2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State
![Page 22: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/22.jpg)
Solution: Identify all Combinations of Consistent “Unknown” Modes
• Candidate: Assignment of G or U to each component.• Diagnosis: Candidate consistent with model and observations.
And(i): G(i):
Out(i) = In1(i) AND In2(i) U(i): No Constraint
Diagnosis = {A1=G, A2=U, A3=G, X1=G, X2=U}
A
B
C
D
E
1
1
1
0
1
F
G
X
Y
Z
0
1
A1
A3
X1
1
0
1
22
2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State
![Page 23: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/23.jpg)
Mode EstimationGiven: Mode, State, Observation Variables: X, Y, and O Y Obs = assignment to O Model: Φ(X,Y) = components + structure
And(i):G(i):
Out(i) = In1(i) AND In2(i)U(i): No Constraint
• All behaviors are associated with modes.• All components have “unknown Mode” U, whose assignment is never mentioned in any constraint.
A
B
C
D
E
1
1
1
0
1
F
G
X
Y
Z
0
1
A1
A3
1
X1
0
1
Return: All mode estimates
2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State
M,obs {X DX | Obs(X,Y ) is satisfiable}
23
![Page 24: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/24.jpg)
Models in Propositional State Logic
And(i): G(i):
Out(i) = In1(i) AND In2(i) i=G {[In1(i)=1 ∧ In2(i)=1] iff Out(i)=1} U(i): No Constraint
Or(i): G(i):
Out(i) = In1(i) OR In2(i) i=G {[In1(i)=1 In2(i)=1] iff Out(i)=1} U(i): No Constraint
X {1,0} X=1 X=0 (i=G) (In1(i)=1) Out(i)=1[X=1 ∧ X=0] (i=G) (In2(i)=1) Out(i)=1
(i=G) (In1(i)=0) (In2(i)=0) Out(i)=0
2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State 24
![Page 25: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/25.jpg)
Outline
• Programs that monitor and control hidden states.
• Consistency-based Diagnosis– Encoding diagnoses compactly using kernels.– Using conflicts to divide and conquer.
25
2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State
![Page 26: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/26.jpg)
Need Compact Encoding
And(i):G(i):
Out(i) = In1(i) AND In2(i)U(i): No Constraint
A
B
C
D G0 11E Z1 A3
1
1
1
FX
Y
0A1
X1
1
0
D,obs {X DX |Y DX st Obs(X,Y)}
As more constraints are relaxed, candidates are more easily satisfied.Typically an exponential number of diagnoses (mode estimates).
How do we encode solutions compactly?
2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State
26
![Page 27: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/27.jpg)
Partial Diagnosis
{A1=U, A2=U, X2=U}
Partial Diagnoses
Partial Diagnosis:
A partial mode assignment M, all of whose full extensions are diagnoses.
• M “removes all symptoms.”
?
?
A
B
C
D
E
1
1
1
0
1
F
G
X
Y
Z
0
1
A3
X1
1
0
1
Diagnoses with common assignments:
{A1=U, A2=U, A3=G, X1=G, X2=U}
{A1=U, A2=U, A3=G, X1=U, X2=U}
{A1=U, A2=U, A3=U, X1=G, X2=U}
{A1=U, A2=U, A3=U, X1=U, X2=U}
27
2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State
![Page 28: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/28.jpg)
Kernel Diagnosis
{A2=U, X2=U}
Kernel Diagnoses
Partial Diagnosis:
A partial mode assignment M, all of whose full extensions are diagnoses.
Kernel Diagnosis:
The smallest partial diagnoses.
A partial diagnosis K, no subset of which is a partial diagnosis.
?
?
?
A
B
C
D
E
1
1
1
0
1
F
G
X
Y
Z
0
1
A1
A3
X1
1
0
1
Partial Diagnosis
{A1=U, A2=U, X2=U}
28
2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State
![Page 29: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/29.jpg)
Outline
• Programs that monitor and control hidden states.
• Consistency-based Diagnosis– Encoding diagnoses compactly using kernels.– Using conflicts to divide and conquer.
29
2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State
![Page 30: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/30.jpg)
Conflicts Explain How to Remove SymptomsA
B
C
D
E
1
1
1
0
F
G
X
Y
Z
Symptom: F is observed 0, but predicted to be 1 if A1, A2 and X1 are okay.
Conflict 1: {A1=G, A2=G, X1=G} is inconsistent.
Conflict: An inconsistent partial assignment to mode variables X.
F 0
11
0
→One of A1, A2 or X1 must be broken.
1
A1
A2
A3
X1
X2
Symptom
302/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State
![Page 31: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/31.jpg)
Second Conflict
Symptom: G is observed 1, but predicted 0.
Conflict 2: {A1=G, A3=G, X1=G, X2=G} is inconsistent.
Symptom
1
1
A
B
C
D
E
F
G
X
Y
Z
1
1
1
0
1
0
11
0
A1
A2
A3
X1
X2
Conflicting modes aren’t always
upstream from symptom.
→One of A1, A3, X1 or X2 must be broken.
31
2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State
![Page 32: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/32.jpg)
Summary: Conflicts1 Symptom
Conflict: A partial mode assignment M that isinconsistent with the model and observations.
Properties:• Every superset of a conflict is a conflict.• Only need conflicts that are minimal under subset.• implies
1
0
A
BCD
E
F
G
X
Y
Z
1
110
1
0
11
1
A1
A2
A3
X1
X2
Obs2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State
M
32
![Page 33: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/33.jpg)
Diagnosis by Divide and Conquer
Given model Φ and observations Obs,1. Find all symptoms.2. Diagnose each symptom separately
(each generates a conflict).3. Merge diagnoses
(set covering → kernel diagnoses).
General Diagnostic Engine[de Kleer & Williams, 87]
Conflict Recognition
CandidateGeneration
33
2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State
![Page 34: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/34.jpg)
Summary: Mode EstimationGiven: Mode, State, Observation Variables: X, Y, and O Y Obs = an assignment to O Model: Φ(X,Y) = components + structure
And(i):G(i):
Out(i) = In1(i) AND In2(i)U(i): No Constraint
• All behaviors are associated with modes.• All components have “unknown Mode” U, whose assignment is never mentioned in any constraint.
A
B
C
D
E
1
1
1
0
1
F
G
X
Y
Z
0
1
A1
A3
X1
1
0
1
M,obs {X DX | Obs(X,Y ) is satisfiabReturn: All mode estimates
le}
2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State
34
![Page 35: Programs that Monitor Hidden State: Mode Estimation and ... · 2/8/16 16.412J/6.834J S16: L2 Monitoring Hidden State . M , obs {X ...](https://reader034.fdocuments.in/reader034/viewer/2022042218/5ec3cd3f6858954c363a33e2/html5/thumbnails/35.jpg)
MIT OpenCourseWarehttps://ocw.mit.edu
16.412J / 6.834J Cognitive RoboticsSpring 2016
For information about citing these materials or our Terms of Use, visit: https://ocw.mit.edu/terms.