Incremental Contingency Planning Richard Dearden, Nicolas Meuleau, Sailesh Ramakrishnan, David E' Sm - PowerPoint PPT Presentation

1 / 40
About This Presentation
Title:

Incremental Contingency Planning Richard Dearden, Nicolas Meuleau, Sailesh Ramakrishnan, David E' Sm

Description:

Drive (-1) Dig(5) Visual servo (.2, -.15) NIR. Lo res. Rock ... Drive(-1) NIR. Compress. Drive(2) Maximize (Expected) Scientific Return. Given: start time ... – PowerPoint PPT presentation

Number of Views:29
Avg rating:3.0/5.0
Slides: 41
Provided by: DavidE130
Category:

less

Transcript and Presenter's Notes

Title: Incremental Contingency Planning Richard Dearden, Nicolas Meuleau, Sailesh Ramakrishnan, David E' Sm


1
Incremental Contingency Planning Richard
Dearden, Nicolas Meuleau, Sailesh Ramakrishnan,
David E. Smith, Rich Washington
?
2
Why Contingency Planning ??
Limited onboard processing CPU, memory,
time Safety sequence checking Anticipation setup
steps
3
The Planning Problem
Given start time pose energy available actions
with uncertain durations resource usage Possible
science objectives images samples
Compress

Visual servo (.2, -.15)
Dig(5)
Drive(-1)
NIR
Drive(2)
Warmup NIR
Maximize (Expected) Scientific Return
4
Technical Challenges
Continuous time ( resources) Continuous
outcomes Time ( resource) constraints Concurren
cy Goal selection optimization g1, g2, g3, g4

5
Just in Case (JIC) Scheduling
1. Seed schedule 2. Identify most likely
failure 3. Generate a contingency
branch 4. Integrate the branch
.4
.2
.1
Advantages Tractability Simple plans Anytime
6
Just in Case (JIC) Planning
1. Seed plan 2. Identify most likely
failure 3. Generate a contingency
branch 4. Integrate the branch
.4
.2
.1
7
Limits of JIC Scheduling Heuristics
t ? 900, 1430 ? 5s ? 1s
? most probable failures most
interesting branch point
Most probable failure points may not be the best
branch-points It is often too late to attempt
other goals when the plan is about to fail.
V 10
HiRes
t ? 1000, 1400 ? 600s ? 60s
? 300s ? 5s
? 120s ? 60s
? 1000s ? 500s

?
?
?
V 100
Dig(60)
Visual servo (.2, -.15)
Drive (-2)
NIR
t ? 900, 1600 ? 5s ? 1s
t ? 1000, 1350 ? 600s ? 60s
? 120s ? 20s
V 50
Lo res
Rock finder
LIB
V 5
Warmup LIB
True for all initial states in the grey box.
? 1200s ? 20s
8
Just in Case (JIC) Planning
?
?
?
?
1. Seed plan 2. Identify best branch
point 3. Generate a contingency branch 5.
Evaluate Integrate the branch
Construct plangraph Back-propagate value
tables Compute gain Select branch condition
goals
Vm
Vb
r
9
Construct Plangraph
g1
g2
g3
g4
10
Value Tables
V1
g1
r
V2
g2
r
V3
g3
r
V4
g4
r
11
Example
q
g
A
B
(10, 15)
(10, 15)
p
C
r
(3, 3)
E
g
(2, 2)
D
s
t
(1, 5)
12
Simple Propagation
q
g
A
B
(10, 15)
(10, 15)
p
C
r
(3, 3)
E
g
(2, 2)
D
s
t
(1, 5)
13
Simple Propagation
q
g
A
B
(10, 15)
(10, 15)
p
C
r
(3, 3)
E
g
(2, 2)
D
s
t
(1, 5)
14
Simple Propagation
q
g
A
B
(10, 15)
(10, 15)
p
C
r
(3, 3)
E
g
(2, 2)
D
s
t
(1, 5)
15
Conjunctions
q
g
A
B
(10, 15)
(10, 15)
p
t
1
C
r
2
(3, 3)
E
g
(2, 2)
D
s
t
(1, 5)
16
Propagation
q
g
A
B
(10, 15)
(10, 15)
p
t
1
C
r
2
(3, 3)
E
g
(2, 2)
D
s
t
(1, 5)
17
Propagation
q
g
A
B
(10, 15)
(10, 15)
p
C
r
(3, 3)
E
g
(2, 2)
D
s
t
(1, 5)
18
Combining Tables
q
g
A
B
(10, 15)
(10, 15)
p
C
r
(3, 3)
E
g
(2, 2)
D
s
t
(1, 5)
19
Discharging Assumptions
q
g
A
B
(10, 15)
(10, 15)
p
C
r
(3, 3)
E
g
(2, 2)
D
s
t
(1, 5)
20
Propagation
q
g
A
B
(10, 15)
(10, 15)
p
C
r
(3, 3)
E
g
(2, 2)
D
s
t
(1, 5)
21
Combining Tables
q
g
A
B
(10, 15)
(10, 15)
p
1
C
r
5
(3, 3)
E
g
(2, 2)
D
s
t
(1, 5)
22
Combining Tables
q
g
A
B
5
(10, 15)
(10, 15)
p
25
8
1
C
r
5
(3, 3)
E
g
(2, 2)
D
s
t
(1, 5)
23
Ordering
q
g
A
B
AB
5
(10, 15)
CDE
(10, 15)
p
25
8
1
C
r
5
(3, 3)
E
g
(2, 2)
D
s
t
DCE
(1, 5)
24
Achieving Multiple Goals
gg
g

5
g
25
15
30
q
g
A
B
5
(10, 15)
(10, 15)
p
25
8
1
C
r
5
(3, 3)
E
g
(2, 2)
D
s
t
(1, 5)
25
Achieving Multiple Goals
q
g
A
B
(10, 15)
(10, 15)
p
1
C
r
5
(3, 3)
E
g
(2, 2)
D
s
t
(1, 5)
26
Goal Annotation
g
g
q
g
A
B
(10, 15)
(10, 15)
p
g
1
C
r
5
g
(3, 3)
E
g
(2, 2)
D
s
t
g
g
(1, 5)
g
27
Just in Case (JIC) Planning
?
?
?
?
1. Seed plan 2. Identify best branch
point 3. Generate a contingency branch 5.
Evaluate Integrate the branch
Construct plangraph Back-propagate value
tables Compute gain Select branch condition
goals
Vm
Vb
r
28
Estimating Branch Value
V1
V
r
Max
V2
V
r
V3
V
r
V4
29
Plan Statistics
resource probability
plan value function
P
Vm
r
r
V1
V2
V3
V4
30
Expected Branch Gain
P
r
V1
V2
V3
V4
31
Selecting the Branch Condition
branch condition
branch condition
P
r
V1
V2
V3
V4
32
Selecting Branch Goals
branch goals
g1
P
g3
r
r
V1
g1
V2
g3
V3
V4
33
Evaluating the Branch
?
?
?
?
1. Seed plan 2. Identify best branch
point 3. Generate a contingency
branch 4. Evaluate integrate the branch
Compute value function Compute actual gain
34
Actual Branch Gain
P
actual branch condition
Vm
Vb
r
r
Branch value function
35
Remarks Single Plangraph
?
?
?
?
1. Seed plan 2. Identify best branch
point 3. Generate a contingency branch 5.
Evaluate Integrate the branch
Construct plangraph Back-propagate value
tables Compute gain Select branch condition
goals
Vm
Vb
r
36
Plan Graph
V1
g1
r
V2
g2
r
V3
g3
r
V4
g4
r
37
Branch Initial Conditions
p
v
r
q,r
v
r
v
r
p,r
v
v
r
r
38
Single Plangraph
Construct single plangraph Back-propagate value
tables
1. Seed plan 2. Identify best branch
point 3. Generate a contingency
branch 4. Evaluate integrate the branch
?
?
?
?
Discharge conditions Compute gain Select branch
condition goals
39
Issues
Sensor costs Setup steps Utility updates Floating
contingencies
40
The End.
Write a Comment
User Comments (0)
About PowerShow.com