Cassie as a Self-Aware SNePS/GLAIR Agent - PowerPoint PPT Presentation

About This Presentation
Title:

Cassie as a Self-Aware SNePS/GLAIR Agent

Description:

Cassie as a. Self-Aware SNePS/GLAIR Agent. Stuart C. Shapiro ... I = SNePS term denoting Cassie. NOW = SNePS term denoting current time ... – PowerPoint PPT presentation

Number of Views:157
Avg rating:3.0/5.0
Slides: 21
Provided by: StuartC93
Learn more at: https://cse.buffalo.edu
Category:
Tags: glair | sneps | agent | aware | cassie | self

less

Transcript and Presenter's Notes

Title: Cassie as a Self-Aware SNePS/GLAIR Agent


1
Cassie as aSelf-Aware SNePS/GLAIR Agent
  • Stuart C. Shapiro
  • Department of Computer Science and Engineering,
  • and Center for Cognitive Science
  • University at Buffalo, The State University of
    New York
  • 201 Bell Hall, Buffalo, NY 14260-2000
  • shapiro_at_cse.buffalo.edu
  • http//www.cse.buffalo.edu/shapiro/

2
What is it?
  • A computational cognitive agent that can
  • Understand and communicate in English
  • Discuss specific, generic, and rule-like
    information
  • Reason
  • Discuss acts and plans
  • Sense
  • Act
  • Maintain a model of itself
  • Remember and report what it has sensed and done.

3
Cassie, the FEVAHR
4
Simulated FEVAHR
Cassie
5
UXO Remediation
Corner flag
Field
Drop-off zone
UXO
NonUXO object
Battery meter
Corner flag
Corner flag
Cassie
Recharging Station
Safe zone
6
Crystal Cassie
First Person Perspective Views
7
Princess from The Trial, The Trail
A VR drama by Josephine Anstey
8
Patofil and Filopat
9
Magellan ProTM Mobile RobotfromiRobot
10
GLAIR Architecture
Grounded Layered Architecture with Integrated
Reasoning
Knowledge Level
Mind
SNePS
Perceptuo-Motor Level
deictic modality registers primitive-action
procedures
B o d y
IP Sockets
NL
Sensory-Actuator Level
Vision
Sonar
World
Motion
Proprioception
11
SNePS
  • A logic-based KR system
  • using reified propositions
  • with special constructs for acts
  • and an acting executive
  • as well as an inference engine
  • with an ATMS.
  • From a Wumpus World agent
  • all(r)(Isa(r,room)
  • gt whendo(In(r), believe(Contains(r,p
    it)))).

12
Kinds of Self-Awarenessin Cassie
  • Static beliefs (non-fluents) about herself
  • Knows and recognizes her name.
  • My name is Cassie.
  • Has first-person, privileged knowledge of ,
  • remembers, and can tell what shes done.
  • Has a sense of time, including now.
  • Knows whom she is conversing with.
  • I talked to Bill and I am talking to you.
  • Has self-perception of what she says.

13
Basic Approach toEssential Self-Awareness
  • Beliefs about self at KL (mind)
  • inserted by PML (body).

14
How Its Implementedat the KL
  • SNePS term denoting Cassie
  • Used in all beliefs about herself.
  • SNePS terms denoting times
  • To relate acts/events
  • Representation of acts is the same
  • For acts to be done by Cassie
  • For acts done by Cassie
  • For acts done by other agents
  • For acting
  • For NL interaction

15
How Its Implementedat the PML
  • Deictic Registers
  • I SNePS term denoting Cassie
  • NOW SNePS term denoting current time
  • YOU SNePS term denoting conversational
    partner

16
How Its Implementedat the PML
  • Modality Registers
  • For current acts
  • Vision modality register
  • KL term of entity Cassies currently looking at

17
How Its Implementedat the PML
  • Lisp implementation of KL primitive acts
  • Implementation of act a inserts KL belief that
  • I did a NOW
  • and moves NOW.
  • First person, privileged belief.

18
Multi-Computer VR Agents
KL
SNePS terms
primitive-action procedures
PMLa
deictic modality registers
PMLb
IP Sockets
PMLc
SAL
hardware drivers
19
Body-Mind FeedbackMost Difficult Issue
  • When should mind move to next step of plan?
  • When sense that goal of previous step has been
    achieved.
  • For speech sequence
  • when hear self over self-perception socket.

20
Next Steps
  • Since awareness of acting is
  • added by implementation of primitive acts,
  • dont yet have awareness of non-primitive
    acts.
  • Put all this in architecture,
  • so agent implementers neednt be concerned.
Write a Comment
User Comments (0)
About PowerShow.com