Brian C. Williams
16.410/16.413
September 3
rd
, 2003
Introduction to Principles of
Autonomy and Decision Making
1
Brian C. Williams
16.410/16.413
September 3
rd
, 2003
Today’s Assignment
?
?
Read Chapters 1 and 2 of AIMA
–
–
“Artificial Intelligence: A Modern Approach”
by Stuart Russell and Peter Norvig
–2
nd
Edition (not 1
st
Edition!!)
nd st
–
– AIMA is available at the Coop
AIMA is available at the Coop
1
Outline
?
? The promise of autonomous explorers
?
? The challenge of autonomous explorers
?
? Agents great and small
?
? Course objective 1 (16.410/13):
–
– Principles for Building Agents
?
?C
C
ourse objective 2 (16.413):
– B
uilding an Agent:
Agent:
–B
The Mars exploration rover (MER) project.
2
.
Courtesy NASA/JPL-Caltech. http://www.jpl.nasa.gov
A
Image taken from NASA's website: http://www.nasa.gov Courtesy of Kanna Rajan.
Frontier . . .
spacecraft.’’
establish a virtual presence, in space, on planets, in aircraft and
``Our vision in NASA is to open the Space We must
- Daniel S. Goldin, NASA Administrator, May 29, 1996
3
Motive: Follow the water ….
? to find evidence of past life on Mars.
? to study the forces that shaped Mars.
? to develop future life on Mars.
Courtesy NASA/JPL-Caltech. http://www.jpl.nasa.gov Courtesy of Kanna Rajan.
Mars
Motive: Follow the water ….
? to find evidence of past life on Mars.
? to study the forces that shaped Mars.
? to develop future life on Mars.
Courtesy U.S. Geological Survey.
4
Inner and Outer Planets Missions
MESSENGER
mission to Mercury
MESSENGER
ury
Venus
Sample Return
Venus
rn
Comet Nucleus
Sample Return
Primitive Bodies Missions
Pluto/Kuiper Express
Europa Orbiter
Europa
Lander
Neptune Orbiter
Titan
Explorer
Cryobot &
Hydrobot
Motive: life underMotive: life under Europa?
Courtesy NASA/JPL-Caltech. http://www.jpl.nasa.gov Courtesy of Kanna Rajan.
5
MIT SPHERES
Image taken from NASA's website. http://www.nasa.gov. Adapted. Courtesy of Kanna Rajan.
6
Cooperative Exploration
Distributed Planning Group, JPL
Model-based Embedded
and Robotic Systems Group, MIT
Courtesy NASA/JPL-Caltech. http://www.jpl.nasa.gov Courtesy of Kanna Rajan.
Outline
?
? The promise of autonomous explorers
?
? The challenge of autonomous explorers
The challenge of autonomous explorers
?
?
Agents great and small
Agents great and small
?
? Course objective 1 (16.410/13):
–
– Principles for Building Agents
?
? Course objective 2 (16.413):
Course objective 2 (16.413):
– Building n Agent:
Building an Agent:
The Mars exploration rover (MER) project.
7
A Capable Robotic Explorer: Cassini
? 7 year cruise
Faster, Better, Cheaper
? ~ 150 - 300
ground
operators
?~ 1 billion $ ?150 million $
? 7 years to ?2 year build
build
? 0 ground
ops
Four launches in 7 months
Mars Climate Orbiter: 12/11/98
/99
Stardust: 2/7/99
QuickSCAT: 6/19/98
Mars Polar Lander: 1/3
courtesy of JPL
8
Mars Polar Lander
L
Spacecraft require a good physical commonsense…
Launch: 1/3/99
Traditional spacecraft commanding
2;
6;
,
Whats a better paradigm?
GS,SITURN,490UA,BOTH,96-355/03:42:00.000;
CMD,7GYON, 490UA412A4A,BOTH, 96-355/03:47:00:000, ON;
CMD,7MODE, 490UA412A4B,BOTH, 96-355/03:47:02:000, INT;
CMD,6SVPM, 490UA412A6A,BOTH, 96-355/03:48:30:000,
CMD,7ALRT, 490UA412A4C,BOTH, 96-355/03:50:32:000,
CMD,7SAFE, 490UA412A4D,BOTH, 96-355/03:52:00:000, UNSTOW;
CMD,6ASSAN, 490UA412A6B,BOTH, 96-355/03:56:08:000, GV,153,IMM,231,
GV,153;
CMD,7VECT, 490UA412A4E,BOTH, 96-355/03:56:10.000, 0,191.5,6.5,
0.0,0.0,0.0,
96-350/
00:00:00.000,MVR;
SEB,SCTEST, 490UA412A23A,BOTH, 96-355/03:56:12.000 SYS1,NPERR;
CMD,7TURN, 490UA412A4F,BOTH, 96-355/03:56:14.000, 1,MVR;
MISC,NOTE, 490UA412A99A,, 96-355/04:00:00.000, ,START OF TURN;,
CMD,7STAR, 490UA412A406A4A,BOTH 96-355/04:00:02.000, 7,1701,
278.813999,38.74;
CMD,7STAR, 490UA412A406A4B,BOTH, 96-355/04:00:04.000, 8,350,120.455999,
-39.8612;
CMD,7STAR, 490UA412A406A4C,BOTH, 96-355/04:00:06.000, 9,875,114.162,
5.341;
CMD,7STAR, 490UA412A406A4D,BOTH, 96-355/04:00:08.000, 10,159,27.239,
89.028999;
CMD,7STAR, 490UA412A406A4E,BOTH, 96-355/04:00:10.000, 11,0,0.0,0.0;
CMD,7STAR, 490UA412A406A4F,BOTH, 96-355/04:00:12.000, 21,0,0.0,0.0;
9
Outline
?
?
The promise of autonomous explorers
?
?
The challenge of autonomous explorers
?
?
Agents great and small
?
?
Course objective 1 (16.410/13):
–
–
Principles for Building Agents
?
?
Course objective 2 (16.413):
–
–
Building an Agent:
Agent:
The Mars exploration rover (MER) project.
Agents and Intelligence
Adaoted from J. Malik, U.C. Berkeley
10
Reflex agents
Adaoted from J. Malik, U.C. Berkeley
Goal-oriented agent
Adaoted from J. Malik, U.C. Berkeley
11
Utility-based agent
Adaoted from J. Malik, U.C. Berkeley
Outline
?
?
The promise of autonomous explorers
?
?
The challenge of autonomous explorers
?
?
Agents great and small
?
?
Course objective 1 (16.410/13):
–
–
Principles for Building Agents
?
?
Course objective 2 (16.413):
–
–
Building an Agent:
Agent:
The Mars exploration rover (MER) project.
12
Course Objective 1:
Course Objective 1:
Principles of Agents
16.410/13: To learn the modeling and
16.410/13: To learn the modeling and
algorithmic building blocks for creating
algorithmic building blocks for creating
reasoning, learning agents:
?
?
To formulate reasoning problems.
?
?
To describe, analyze and demonstrate
o describe, analyze and demonstrate
reasoning algorithms.
?
?
To model and encode
o model and encode
knowledge used by
knowledge used by
reasoning algorithms.
Agent Paradigms
? Extensive Reasoning
? Extensive Learning
? Extensive Optimization
13
Extensive Reasoning:
Houston, we have a problem ...
courtesy of NASA
? Quintuple fault occurs
(three shorts, tank-line and
pressure jacket burst, panel
flies off).
? Mattingly works in ground
simulator to identify new
sequence handling severe
power limitations.
? Mattingly identifies novel
reconfiguration, exploiting
LEM batteries for power.
? Swaggert & Lovell work on
Apollo 13 emergency rig
lithium hydroxide unit.
Image taken from NASA's website: http://www.nasa.gov. Courtesy of Kanna Rajan.
Example of a Model-based Agent:
? Goal-directed
? First time correct
? projective
? reactive
? Commonsense models
? Heavily deductive
Scripts
component models
Goals
Diagnosis
& Repair
Mission
Manager
Executive
Planner/
Scheduler
Remote Agent
Mission-level
actions &
resources
14
Reasoning methods that
Reasoning methods that
support the creation of agents
?
?
Rule-b
ased systems
–
–
Forwa
rd chaining
–
–
Goal
-directed
directed-
?
?
Search and roadm
ap path planning
–
–
Blind Search
Blind Search
–
–
Inform
ed Search
–
–
Adversa
rial (Game) Search
?
?
Planning and Acting in the W
orld
?
?
Constraints and Scheduling
?
?
Model-based Diagnosis
?
?
Logic and Deduction
Extensive Learning:
TD-Gammon [Tesauro, 1995]
TD-Gammon [Tesauro, 1995]
Learns to play Backgammon
Situations:
? Board configurations (10
20
)
Actions:
? Moves
Rewards:
– +100 if win
– - 100 if lose
– 0 for all other states
? Trained by playing 1.5 million games against self.
? Currently, roughly equal to best human player.
t n
s:
ard s )
Actions:
s
Rewards:
– +100 if win
lose
– all other states
? ned by f.
15
Learning methods that support
Learning methods that support
the creation of agents
?
?
Learning through reinforcement
earning through reinforcement
?
?
Learning decision trees
?
?
Neural net learning
Extensive Optimization
16
Cooperative Path Planning:
MILP Encoding: Fuel Equation
Cooperative Path Planning:
MILP Encoding: Fuel Equation
min = J
T
= min 6 q’w
i
+ 6 r’v
i
+ p’w
N
min = J
T i
+
i
w
i
, v
i
w
i
, v
i
i=1
N-1
i=1
N-1
slack control vector weighting vectors
slack state vector
past-horizon
terminal cost term
total fuel calculated over all time
instants i
Cooperative Path Planning:
MILP Encoding: Constraints
?s
ij
ij
<= w
ij
, etc.
ij
State Space Constraints
?
? s
i
i
+1 = As
i
i
+ Bu
+
i
i
State Evolution Equation
-x
?x
i
i
<= x
min
+ Mt
i1
n
i
i
<= -x + Mt
i2max
-y
y
i
i
<= y
min
+ Mt
i3
n
Obstacle Avoidance
i
i
<= -y
max
+ Mt
i4
(for all time i)
6 t
ik
<= 3 (t introduce IP element)
?
Sim
ilar equation for Collision Avoidance (for all pairs of
voidance (for all pairs of
vehicles)
17
Outline
?
?T
T
he promise of autonomous explorers
?
?T
T
he challenge of autonomous explorers
?
?A
A
gents great and small
?
?C
C
ourse objective 1 (16.410/13):
–
– Principles for Building Agents
?
?C
C
ourse objective 2 (16.413):
–
–B
B
uilding an Agent:
Agent:
The Mars exploration rover (MER) project.
Optimization methods that
Optimization methods that
support the creation of agents
Modeling Frameworks:
– Markov Decision Processes
Markov Decision Processes
Mixed Integer Linear Programming
–
–
Mixed Integer Linear Programm
Solution Methods:
– Dyna
Dynamic Programming
–Simplex
– Simplex
– Branch and Bound
Branch and Bound
18
Course Objective 2:
Course Objective 2:
Building Agents
16.413: To appreciate the challenges of building a
16.413: To appreciate the challenges of building a
state of the art autonomous explorer:
?
? To model and encode knowledge needed to solve
To m
a state of the art challenge.
?
? To work through the
To work through the
process of autonomy systems
process of autonomy system
integration.
?
? To assess the promise, frustrations and challenges
To assess the prom
of using (b)leading art technologies.
Mars Exploration Rovers – Jan. 2004
Mars Exploration Rovers – Jan. 2004
Mission Objectives:
? Learn about ancient water and climate on Mars.
? For each rover, analyze a total of 6-12 targets
– Targets = natural rocks, abraded rocks, and soil
? Drive 200-1000 meters per rover
? Take 1-3 panoramas both with Pancam and mini-TES
? Take 5-15 daytime and 1-3 nightime sky observations with
mini-TES
Mission Objectives:
ate on Mars.
alyze a total of 6-12 targets
ts il
ters per rover
as both with Pancam ni-TES
ni-TES
Mini-TES
Pancam
Navcam
Rock Abrasion Tool
Microscopic Imager
Mossbauer spectrometer
APXS
Courtesy NASA/JPL-Caltech. http://www.jpl.nasa.gov. Courtesy of Kanna Rajan.
19
Surface Operations Scenario
Day 4
Day 1
Day 2
Board Navigation
Changes, as needed
Day 3
Mars Exploration Rover
Target
During the Day
Science Activities
Long-Distance Traverse
(<20-50 meters)
Initial Position;
Followed by
“Close Approach”
During the Day
Autonomous On-
Day 2 Traverse Estimated
Error Circle
Science Prep
(if Required)
Day 2 Traverse
Estimated Error Circle
Courtesy of Kanna Rajan, NASA Ames. Used with permission.
Activity Name
Durati
on
10 11 12 13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9
10 11 12 13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9
DTE
4.50
0.75
DTE period DFE
Night Time Rover Operations
16.97
Night Time Rover OperationsSleep Wakeup
Pre-Comm Session Sequence Plan Review
Current Sol Sequence Plan Review
1.50
1.50
Current Sol Sequence Plan Review
Prior Sol Sequence Plan Review
2.00
Prior Sol Sequence Plan Review
Real-TIme Monitoring
4.50
0.75
Real-TIme Monitoring Real-TIme Monitoring
2.75
Downlink Product Generation
Tactical Science
Assessment/Observation Planning
5.00
Tactical Science Assessment/Observation Planning
Science DL Assessment Meeting
1.00
Science DL Assessment Meeting
Payload DL/UL Handoffs
0.50
Payload DL/UL Handoffs
Tactical End-of-Sol Engr. Assessment &
Planning
5.50
Tactical End-of-Sol Engr. Assessment & Planning
DL/UL Handover Meeting
0.50
DL/UL Handover Meeting
Skeleton Activity Plan Update
2.50
Skeleton Activity Plan Update
SOWG Meeting
2.00
SOWG Meeting
Uplink Kickoff Meeting
0.25
Uplink Kickoff Meeting
Activity Plan Integration & Validation
1.75
Activity Plan Integration & Validation
Activity Plan Approval Meeting
0.50
Activity Plan Approval Meeting
Build & Validate Sequences
2.25
Build & Validate Sequences
UL1/UL2 Handover
1.00
UL1/UL2 Handover
Complete/Rework Sequences
2.50
Complete/Rework Sequences
Margin 1
0.75
Margin 1
Command & Radiation Approval
0.50
Command & Radiation Ap
Margin 2
1.25
Margin 2
Radiation
0.50
Radiation
MCT Team
7.00
4.00
One day in the life of a Mars rover
Science Planning Sequence Build/Validation Uplink
Downlink Product Generation...
Courtesy: Jim Erickson
Downlink Assessment
Courtesy of Kanna Rajan, NASA Ames. Used with permission.
20
EUROPA
Automated
Planning System
EUROPA
Automated
Planning System
Science
Navigation
Engineering
Resource
Constraints
DSN/Telcom
Flight Rules
Science Team
Sequence
Build
MAPGEN: Automated
Science Planning for MER
Planning Lead: Kanna Rajan (ARC)
Courtesy of Kanna Rajan, NASA Ames. Used with permission.
Course Challenge
? What would it be like to operate MER if it
was fully autonomous?
Course project:
? Demonstrate an autonomous MER mission
in simulation, and in the MIT rover testbed.
was fully autonomous?
in simulation, and in the MIT rover testbed.
21
Next Challenge: Mars Smart Lander
(2009)
(2009)
Mission Duration: 1000 days
Total Traverse: 3000-69000 meters
Meters/Day: 230-450
Science Mission: 7 instruments, sub-surface science
package (drill, radar), in-situ sample “lab”
Technology Demonstration:
(2005).
Courtesy NASA/JPL-Caltech. http://www.jpl.nasa.gov. Courtesy of Kanna Rajan.
22