CS344M Autonomous Multiagent Systems Todd Hester Department or - - PowerPoint PPT Presentation

▶

Jan 29, 2023 285 likes •611 views

CS344M Autonomous Multiagent Systems Todd Hester Department or Computer Science The University of Texas at Austin Good Afternoon, Colleagues Todd Hester Good Afternoon, Colleagues Are there any questions? Todd Hester Logistics Office

SLIDE 1

CS344M Autonomous Multiagent Systems

Todd Hester Department or Computer Science The University of Texas at Austin

SLIDE 2

Good Afternoon, Colleagues

Todd Hester

SLIDE 3

Good Afternoon, Colleagues

Are there any questions?

Todd Hester

SLIDE 4

Logistics

Office hours

Todd Hester

SLIDE 5

Logistics

Office hours
Reading responses

Todd Hester

SLIDE 6

Logistics

Office hours
Reading responses
Resources page

Todd Hester

SLIDE 7

Logistics

Office hours
Reading responses
Resources page
Programming assignment

− How’s it going?

Todd Hester

SLIDE 8

Logistics

Office hours
Reading responses
Resources page
Programming assignment

− How’s it going? − Important: 3D code is only for this class

Todd Hester

SLIDE 9

Logistics

Office hours
Reading responses
Resources page
Programming assignment

− How’s it going? − Important: 3D code is only for this class

Talks in the department:

Todd Hester

SLIDE 10

Logistics

Office hours
Reading responses
Resources page
Programming assignment

− How’s it going? − Important: 3D code is only for this class

Talks in the department:

− Jeff Clune, Friday at 11am (2.402) − Cornell University − “Automatically generating regular, modular neural networks with computational abstractions of evolution and developmental biology”

Todd Hester

SLIDE 11

Logistics

Next week’s readings up

− Multiagent Systems – an overview − Another overview (optional) − Pushing Brooks’ approach to MAS

Todd Hester

SLIDE 12

Logistics

Next week’s readings up

− Multiagent Systems – an overview − Another overview (optional) − Pushing Brooks’ approach to MAS − Our RoboCup team paper

Todd Hester

SLIDE 13

Logistics

Next week’s readings up

− Multiagent Systems – an overview − Another overview (optional) − Pushing Brooks’ approach to MAS − Our RoboCup team paper − Free-form response

Todd Hester

SLIDE 14

Logistics

Next week’s readings up

− Multiagent Systems – an overview − Another overview (optional) − Pushing Brooks’ approach to MAS − Our RoboCup team paper − Free-form response

Todd Hester

SLIDE 15

Writing Assessment

What did you think of these readings?
What was good about them?
How could they have improved?

Todd Hester

SLIDE 16

Writing

Direct, articulate responses

− Thesis sentence − Supporting argument − Demonstrate that you know what you’re saying

Todd Hester

SLIDE 17

Writing

Direct, articulate responses

− Thesis sentence − Supporting argument − Demonstrate that you know what you’re saying One way that TCA departs from Rodney Brooks’ design principles is that TCA employs a central control

module. TCA’s central component routes messages to

the various connected modules and maintains control information. Brooks’ designs,

n the other hand,

connected perception directly to actions, bypassing any form of central control and also any central representation of the world.

Todd Hester

SLIDE 18

Reactive vs. deliberative (3 senses)

− Respond in a timely fashion − No complex respresentation − No state at all (respond to current percepts)

Todd Hester

SLIDE 19

Reactive vs. deliberative (3 senses)

− Respond in a timely fashion − No complex respresentation − No state at all (respond to current percepts) From the book:

action : P → A
Decision based entirely on the present

Todd Hester

SLIDE 20

Reactive vs. deliberative (3 senses)

− Respond in a timely fashion − No complex respresentation − No state at all (respond to current percepts) From the book:

action : P → A
Decision based entirely on the present

− True of Brooks’ “reactive” agents?

Todd Hester

SLIDE 21

Brooks’ Goals

Todd Hester

SLIDE 22

Brooks’ Goals

Autonomous mobile agents that are seen as intelligent
No interest in applications
Timely, robust, do something

Todd Hester

SLIDE 23

Brooks’ Goals

Autonomous mobile agents that are seen as intelligent
No interest in applications
Timely, robust, do something
How differ from 3T goals?

Todd Hester

SLIDE 24

Brooks’ Goals

Autonomous mobile agents that are seen as intelligent
No interest in applications
Timely, robust, do something
How differ from 3T goals?

− What are their stances towards modeling biology? − Which is more biologically plausible?

Todd Hester

SLIDE 25

Brooks’ Goals

Autonomous mobile agents that are seen as intelligent
No interest in applications
Timely, robust, do something
How differ from 3T goals?

− What are their stances towards modeling biology? − Which is more biologically plausible? Subsumption Architecture

Todd Hester

SLIDE 26

Brooks’ Goals

Autonomous mobile agents that are seen as intelligent
No interest in applications
Timely, robust, do something
How differ from 3T goals?

− What are their stances towards modeling biology? − Which is more biologically plausible? Subsumption Architecture (journal article, page 2)

Todd Hester

SLIDE 27

Merkwelt

Todd Hester

SLIDE 28

Merkwelt

Merkwelt ∼ “perceptual world”
Every agent has its own Merkwelt.

Todd Hester

SLIDE 29

Merkwelt

Merkwelt ∼ “perceptual world”
Every agent has its own Merkwelt.
Why should robots use a representation based on our

Merkwelt?

Do we know our own Merkwelt?

Todd Hester

SLIDE 30

Modules

“When researchers working on a particular module get to choose both the inputs and the outputs that specify the module requirements I believe there is little chance the work they do will fit into a complete intelligent system.” Does this apply to 3T?

Todd Hester

SLIDE 31

Could the 3T apps have used subsumption?

Why or why not?

Todd Hester