Research at the Boundary of Robotics and AI Prof: Peter Stone - - PowerPoint PPT Presentation

▶
research at the boundary of robotics and ai
SMART_READER_LITE
LIVE PREVIEW

Research at the Boundary of Robotics and AI Prof: Peter Stone - - PowerPoint PPT Presentation

Sep 22, 2022 325 likes •770 views

Research at the Boundary of Robotics and AI Prof: Peter Stone Department of Computer Science The University of Texas at Austin AI and Robotics Challenge problems Peter Stone AI and Robotics Challenge problems Always on robots (in

slide-1
SLIDE 1

Research at the Boundary of Robotics and AI

Prof: Peter Stone Department of Computer Science The University of Texas at Austin

slide-2
SLIDE 2

AI and Robotics

  • Challenge problems

Peter Stone

slide-3
SLIDE 3

AI and Robotics

  • Challenge problems
  • Always on robots (in a human-occupied space)

Peter Stone

slide-4
SLIDE 4

AI and Robotics

  • Challenge problems
  • Always on robots (in a human-occupied space)
  • Ad hoc teamwork

Peter Stone

slide-5
SLIDE 5

AI and Robotics

  • Challenge problems
  • Always on robots (in a human-occupied space)
  • Ad hoc teamwork
  • Our role in a climate where industry is interested

Peter Stone

slide-6
SLIDE 6

A Goal of AI and Robotics

Robust, fully autonomous agents in the real world

Peter Stone

slide-7
SLIDE 7

A Goal of AI and Robotics

Robust, fully autonomous agents in the real world How?

Peter Stone

slide-8
SLIDE 8

A Goal of AI and Robotics

Robust, fully autonomous agents in the real world How?

  • Build complete solutions to relevant challenge tasks

Peter Stone

slide-9
SLIDE 9

A Goal of AI and Robotics

Robust, fully autonomous agents in the real world How?

  • Build complete solutions to relevant challenge tasks
  • Drives research on component algorithms, theory

Peter Stone

slide-10
SLIDE 10

A Goal of AI and Robotics

Robust, fully autonomous agents in the real world How?

  • Build complete solutions to relevant challenge tasks
  • Drives research on component algorithms, theory
  • A top-down, empirical approach

Peter Stone

slide-11
SLIDE 11

A Goal of AI and Robotics

Robust, fully autonomous agents in the real world How?

  • Build complete solutions to relevant challenge tasks
  • Drives research on component algorithms, theory
  • A top-down, empirical approach

“Good problems . . . produce good science” [Cohen, ’04]

Peter Stone

slide-12
SLIDE 12

Bottom-Up Metaphors

Russell, ’95 “Theoreticians can produce the AI equivalent of bricks, beams, and mortar with which AI architects can build the equivalent of cathedrals.”

Peter Stone

slide-13
SLIDE 13

Bottom-Up Metaphors

Russell, ’95 “Theoreticians can produce the AI equivalent of bricks, beams, and mortar with which AI architects can build the equivalent of cathedrals.” Koller, ’01 “In AI . . . we have the tendency to divide a problem into well-defined pieces, and make progress on each one.

Peter Stone

slide-14
SLIDE 14

Bottom-Up Metaphors

Russell, ’95 “Theoreticians can produce the AI equivalent of bricks, beams, and mortar with which AI architects can build the equivalent of cathedrals.” Koller, ’01 “In AI . . . we have the tendency to divide a problem into well-defined pieces, and make progress on each one. . . . Part of our solution to the AI problem must involve building bridges between the pieces.”

Peter Stone

slide-15
SLIDE 15

Dividing the Problem

Vision Multiagent Reasoning Game Theory Learning Robotics Representation Knowledge Distributed Optimization Natural Language

AI

Peter Stone

slide-16
SLIDE 16

The Bricks

Vision Multiagent Reasoning Game Theory Learning Robotics Representation Knowledge Distributed Optimization Natural Language

Peter Stone

slide-17
SLIDE 17

The Beams and Mortar

Vision Multiagent Reasoning Game Theory Learning Robotics Representation Knowledge Distributed Optimization Natural Language

Peter Stone

slide-18
SLIDE 18

Towards a Cathedral? ?

Vision Multiagent Reasoning Game Theory Learning Robotics Representation Knowledge Distributed Optimization Natural Language

Peter Stone

slide-19
SLIDE 19

Or Something Else?

Vision Multiagent Reasoning Game Theory Learning Robotics Representation Knowledge Distributed Optimization Natural Language

?

Peter Stone

slide-20
SLIDE 20

A Different Problem Division AI

Peter Stone

slide-21
SLIDE 21

Top-Down Approach

Vision Multiagent Reasoning Game Theory Learning Robotics Representation Knowledge Distributed Optimization Natural Language

“Good problems . . . produce good science” [Cohen, ’04]

Peter Stone

slide-22
SLIDE 22

Meeting in the Middle

Vision Multiagent Reasoning Game Theory Learning Robotics Representation Knowledge Distributed Optimization Natural Language

Peter Stone

slide-23
SLIDE 23

Meeting in the Middle

Vision Multiagent Reasoning Game Theory Learning Robotics Representation Knowledge Distributed Optimization Natural Language

Top-down approaches underrepresented

Peter Stone

slide-24
SLIDE 24

Choosing the Challenge

  • Features of good challenges: [Cohen, ’04]

− Frequent tests; Graduated series of challenges − Accept poor performance; Complete agents

Peter Stone

slide-25
SLIDE 25

Choosing the Challenge

  • Features of good challenges: [Cohen, ’04]

− Frequent tests; Graduated series of challenges − Accept poor performance; Complete agents

  • Closed loop + specific goal
  • 50-year technical, scientific goals

− Beyond commercial applications — not possible now − Moore’s law not enough

Peter Stone

slide-26
SLIDE 26

Choosing the Challenge

  • Features of good challenges: [Cohen, ’04]

− Frequent tests; Graduated series of challenges − Accept poor performance; Complete agents

  • Closed loop + specific goal
  • 50-year technical, scientific goals

− Beyond commercial applications — not possible now − Moore’s law not enough

  • There are many — choose one that inspires you

Peter Stone

slide-27
SLIDE 27

Good Problems Produce Good Science

Manned flight Apollo mission Manhattan project RoboCup soccer Goal: By the year 2050, a team of humanoid robots that can beat the human World Cup champion team.

[Kitano, ’97]

Peter Stone

slide-28
SLIDE 28

RoboCup Soccer

  • Still in progress
  • Many virtues:

− Incremental challenges, closed loop at each stage − Robot design to multi-robot systems − Relatively easy entry − Inspiring to many

  • Visible progress

Peter Stone

slide-29
SLIDE 29

AI and Robotics

  • Challenge problems
  • Always on robots (in a human-occupied space)
  • Ad hoc teamwork
  • Our role in a climate where industry is interested

Peter Stone

slide-30
SLIDE 30

Teamwork

Peter Stone

slide-31
SLIDE 31

Teamwork

Peter Stone

slide-32
SLIDE 32

Teamwork

  • Typical scenario: pre-coordination

− People practice together − Robots given coordination languages, protocols − “Locker room agreement” [Stone & Veloso, ’99]

Peter Stone

slide-33
SLIDE 33

Ad Hoc Teams

  • Ad hoc team player is an individual

− Unknown teammates (programmed by others)

Peter Stone

slide-34
SLIDE 34

Ad Hoc Teams

  • Ad hoc team player is an individual

− Unknown teammates (programmed by others)

  • Teammates likely sub-optimal: no control

Peter Stone

slide-35
SLIDE 35

Ad Hoc Teams

  • Ad hoc team player is an individual

− Unknown teammates (programmed by others)

  • Teammates likely sub-optimal: no control

Peter Stone

slide-36
SLIDE 36

Ad Hoc Teams

  • Ad hoc team player is an individual

− Unknown teammates (programmed by others)

  • Teammates likely sub-optimal: no control

Goal: Create a good team player

Peter Stone

slide-37
SLIDE 37

Ad Hoc Teams

  • Ad hoc team player is an individual

− Unknown teammates (programmed by others)

  • Teammates likely sub-optimal: no control

Goal: Create a good team player

  • Introduced as AAAI Challenge Problem

[Stone et al. ’10]

Peter Stone

slide-38
SLIDE 38

Ad Hoc Teams

  • Ad hoc team player is an individual

− Unknown teammates (programmed by others)

  • Teammates likely sub-optimal: no control

Goal: Create a good team player

  • Introduced as AAAI Challenge Problem

[Stone et al. ’10]

− Theory: repeated games, bandits

[AIJ, ’11]

Peter Stone

slide-39
SLIDE 39

Ad Hoc Teams

  • Ad hoc team player is an individual

− Unknown teammates (programmed by others)

  • Teammates likely sub-optimal: no control

Goal: Create a good team player

  • Introduced as AAAI Challenge Problem

[Stone et al. ’10]

− Theory: repeated games, bandits

[AIJ, ’11]

− Experiments: pursuit, flocking

[Barrett, Genter, ’12]

Peter Stone

slide-40
SLIDE 40

Ad Hoc Teams

  • Ad hoc team player is an individual

− Unknown teammates (programmed by others)

  • Teammates likely sub-optimal: no control

Goal: Create a good team player

  • Introduced as AAAI Challenge Problem

[Stone et al. ’10]

− Theory: repeated games, bandits

[AIJ, ’11]

− Experiments: pursuit, flocking

[Barrett, Genter, ’12]

− RoboCup experiments;

Peter Stone

slide-41
SLIDE 41

Ad Hoc Teams

  • Ad hoc team player is an individual

− Unknown teammates (programmed by others)

  • Teammates likely sub-optimal: no control

Goal: Create a good team player

  • Introduced as AAAI Challenge Problem

[Stone et al. ’10]

− Theory: repeated games, bandits

[AIJ, ’11]

− Experiments: pursuit, flocking

[Barrett, Genter, ’12]

− RoboCup experiments; AAAI Workshops

Peter Stone

slide-42
SLIDE 42

AI and Robotics

  • Challenge problems
  • Always on robots (in a human-occupied space)
  • Ad hoc teamwork
  • Our role in a climate where industry is interested

Peter Stone