SLIDE 1
Quiz: Making it autonomous
Sensors, Algorithms, Models WHAT IS MISSING?
1 / 29 N
Where is academia’s role?
2 / 29 N
Quiz: Making it autonomous WHAT Sensors, IS Algorithms, MISSING? - - PowerPoint PPT Presentation
Quiz: Making it autonomous WHAT Sensors, IS Algorithms, MISSING? - - PowerPoint PPT Presentation
Quiz: Making it autonomous WHAT Sensors, IS Algorithms, MISSING? Models 1 / 29 N Where is academias role? 2 / 29 N Takeaway Research in bits and pieces Certified autonomous product Outstanding challenge for complex,
SLIDE 2
SLIDE 3
Takeaway
Research in ’bits and pieces’ → Certified autonomous product
Outstanding challenge for complex, autonomous systems
3 / 29 N
SLIDE 4
A holistic approach to systems autonomy
For complex, intelligent, safety-critical systems Sagar Behere behere@kth.se 14 November 2014 ITA, São Jose dos Campos, Brazil
4 / 29 A holistic approach to systems autonomy N
SLIDE 5
Who am I?
Self-driving trucks Powertrain control Self-driving cars Private Pilot Flying machines Outdoor robots Cognitive robots Industrial robots 5 / 29 A holistic approach to systems autonomy N
SLIDE 6
Background: Intelligence and autonomy
Contents
Background Holistic autonomy KTH
6 / 29 A holistic approach to systems autonomy N
SLIDE 7
Background: Intelligence and autonomy
What is Intelligence?
7 / 29 A holistic approach to systems autonomy N
SLIDE 8
Background: Intelligence and autonomy
What is Intelligence?
The ability of a system to act appropriately in an uncertain environment
7 / 29 A holistic approach to systems autonomy N
SLIDE 9
Background: Intelligence and autonomy
What is Intelligence?
The ability of a system to act appropriately in an uncertain environment Appropriate action is that which increases the probability of success
7 / 29 A holistic approach to systems autonomy N
SLIDE 10
Background: Intelligence and autonomy
What is Intelligence?
The ability of a system to act appropriately in an uncertain environment Appropriate action is that which increases the probability of success Success is the achievement of behavioral sub-goals that support the system’s ultimate goal
7 / 29 A holistic approach to systems autonomy N
SLIDE 11
Background: Intelligence and autonomy
What is Intelligence?
The ability of a system to act appropriately in an uncertain environment Appropriate action is that which increases the probability of success Success is the achievement of behavioral sub-goals that support the system’s ultimate goal The criteria of success and the system’s ultimate goal may be defined external to the intelligent system
(source: J. Albus, Outline of a theory of intelligence)
7 / 29 A holistic approach to systems autonomy N
SLIDE 12
Background: Intelligence and autonomy
What is autonomy?
8 / 29 A holistic approach to systems autonomy N
SLIDE 13
Background: Intelligence and autonomy
What is autonomy?
The ability to operate without human supervision/intervention
8 / 29 A holistic approach to systems autonomy N
SLIDE 14
Background: Intelligence and autonomy
Summary: Intelligent autonomy
Intelligent autonomy requires decisional processes Decision: notion of deliberation, planning, prediction and evaluation of the outcomes of an action
9 / 29 A holistic approach to systems autonomy N
SLIDE 15
The holistic approach to autonomy
Holistic
ho·lis·tic /h¯
- "listik/ [adj.]
parts → intimately connected, and understandable → only by reference to the whole
10 / 29 A holistic approach to systems autonomy N
SLIDE 16
The holistic approach to autonomy
Holistic
ho·lis·tic /h¯
- "listik/ [adj.]
parts → intimately connected, and understandable → only by reference to the whole So what is necessary for a holistic approach to autonomy? → A systems engineering perspective
10 / 29 A holistic approach to systems autonomy N
SLIDE 17
The holistic approach to autonomy
Systems engineering concerns
Requirements Specification Design Implementation Testing Verification and validation Safety engineering Standards and certification Maintenance, upgrades
11 / 29 A holistic approach to systems autonomy N
SLIDE 18
The holistic approach to autonomy
From algorithms to systems
Requirements Specification Design Implementation Testing Verification and validation Safety engineering Standards and certification Maintenance, upgrades Architecture Algorithms Methods Perception Decision making Control Search
12 / 29 A holistic approach to systems autonomy N
SLIDE 19
The holistic approach to autonomy
From algorithms to systems
Requirements Specification Design Implementation Testing Verification and validation Safety engineering Standards and certification Maintenance, upgrades Architecture Algorithms Methods Perception Decision making Control Search But will it certify?
13 / 29 A holistic approach to systems autonomy N
SLIDE 20
The holistic approach to autonomy
Development process support
Requirements Specification Design Implementation Testing Verification and validation Safety engineering Standards and certification Maintenance, upgrades Hazard analysis Safety goals
14 / 29 A holistic approach to systems autonomy N
SLIDE 21
The holistic approach to autonomy
Development process support
Requirements Specification Design Implementation Testing Verification and validation Safety engineering Standards and certification Maintenance, upgrades Hazard analysis Safety goals Architecture
15 / 29 A holistic approach to systems autonomy N
SLIDE 22
The holistic approach to autonomy
Development process support
Requirements Specification Design Implementation Testing Verification and validation Safety engineering Standards and certification Maintenance, upgrades Hazard analysis Safety goals Architecture Traceability?
16 / 29 A holistic approach to systems autonomy N
SLIDE 23
The holistic approach to autonomy
Product complexity
17 / 29 A holistic approach to systems autonomy N
SLIDE 24
The holistic approach to autonomy
Product complexity
18 / 29 A holistic approach to systems autonomy N
SLIDE 25
The holistic approach to autonomy
Key message
’Autonomy’ is more than just another requirement Individual sensors, perception, control are all "getting there" but.. ..problems of integrating them into a safe, total system are not even completely understood
Analysis methods are inadequate Laws, regulations, standards are not up to speed Exploding state space, difficulties guaranteeing behavior . . .
19 / 29 A holistic approach to systems autonomy N
SLIDE 26
Research at KTH/Sweden and collaboration
Main research units
Center for Autonomous Systems (CAS) Automatic Control Mechatronics KTH Transport Labs
20 / 29 A holistic approach to systems autonomy N
SLIDE 27
Research at KTH/Sweden and collaboration
CAS - Unmanned aerial vehicles
21 / 29 A holistic approach to systems autonomy N
SLIDE 28
Research at KTH/Sweden and collaboration
CAS - Unmanned ground vehicles
22 / 29 A holistic approach to systems autonomy N
SLIDE 29
Research at KTH/Sweden and collaboration
CAS - Underwater vehicles
23 / 29 A holistic approach to systems autonomy N
SLIDE 30
Research at KTH/Sweden and collaboration
CAS - Indoor robotics
24 / 29 A holistic approach to systems autonomy N
SLIDE 31
Research at KTH/Sweden and collaboration
Automatic control
Situation awareness Task/Mission (re)planning Predictive control under constraints Model identification Path planning
25 / 29 A holistic approach to systems autonomy N
SLIDE 32
Research at KTH/Sweden and collaboration
Mechatronics
26 / 29 A holistic approach to systems autonomy N
SLIDE 33
Research at KTH/Sweden and collaboration
Research Concept Vehicle
27 / 29 A holistic approach to systems autonomy N
SLIDE 34
Research at KTH/Sweden and collaboration
Collaboration opportunities
Exchange
Master students, interns, thesis workers PhD students (typically one semester) Individual researchers
Brazilian Industry
Joint projects Assignments Case studies
28 / 29 A holistic approach to systems autonomy N
SLIDE 35
Research at KTH/Sweden and collaboration
Takeaway
Research in ’bits and pieces’ → Certified autonomous product
Outstanding challenge for complex, autonomous systems
29 / 29 A holistic approach to systems autonomy N