Robots Learning from Humans

Chapter 10. The Explorer System
in Cognitive Systems, Christensen et al.
2nd Part
Course: Robots Learning from Humans
Park, Seong-Beom
Behavioral neurophysiology lab
Brain and Cognitive Science department,
Seoul National University
http://inahlee.org
Table of contents
• Planning
• Scenario: Find Object
• Real navigation of rats
Planning: explorer scenario should include
Goals
• Motivation of
action
beliefs
• prior
knowledge
about object
position and
spatial
relations
actions
• Perceive and
manipulation
of
environment
Planning: External actions and internal
actions
External(physical)
actions
Follow person
Approach person
Gain attention
Move
Object search
Inform
Internal action
(e.g. processing)
Conceptual Mapping subarchitecture
e.g. query difault information, e.g. about where books are usually found.
Scenario: find object
When the robot is ordered by a user to “Find me the Boland book”, What should it do?
Initial Binding
state
• Initialize current
state, subjects
and space
• robot, current
area, person and
so on
Processing
Utterance
• interpret the
meaning of
words
• “Me”, “Boland
book”
Motive Generation
• set the goal of
action
• “(:goal (K
motmon4
(perceivedposition
motmon6)))” in
MAPL, where
motmon4 is
owner and
motmon6 is the
book
Continual Plan
Creation
Plan execution and
revision
• Determine
sequence of
action
• 1.
Acknowledge
command
acceptance
• 2. Think where
the book is.
• Search the
position
• Tell the user
the position of
book.
• Do find the book
Real navigation system
• Hippocampus cells KNOW where I am.
• HOW?
• Two strategy : Cognitive map and path integration
John O’keefe
(Nobel prize winner)
Real Navigation system
Path integration
Cognitive map
• Calculate vector from it’s
path way
• Direction, distance, speed
information was need
• Calculate distance from
anchor point (wall)