4DRCS

1
4D/RCS A Cognitive Systems Architecture for
Full Spectrum Operations
James Albus Senior NIST Fellow Intelligent
Systems Division Manufacturing Engineering
Laboratory
james.albus_at_nist.gov
2
What is a Cognitive System?
A network of intelligent modules
Examples of intelligent modules perception segm
entation grouping classification /
recognition world modeling simulation,
visualization, inferencing decision-making,
planning, control
3
What does a Cognitive System Consist of?

• Functional modules
• Communication between modules (network)
• Declarative knowledge about the world and the
  self
• Procedural knowledge for
• perception, attention
• modeling, simulation
• reasoning
• decision-making,
• planning
• control

4
What is a Network?
Nodes Edges
Nodes gt Routers or functional modules Edges gt
Communication pathways
5
The network communicates messages between the
nodes
What goes on in the nodes?
Message routing only?
Computational processes? Functional
transformations Knowledge in data
structures Perception, World Modeling
Decision-Making, Planning, Control
6
At the Nets Edge

• Unmanned systems will play a major role,
• but teleoperated systems will be limited

• Teleoperation is bandwidth intensive

• Teleoperation imposes a heavy operator
• workload

Large numbers of teleoperated ground systems are
impractical
7
A Reality Check
The battlefield is not the internet
Bandwidth is limited Communications are
unreliable Jamming is a problem Drop outs are
common
Inside buildings In urban canyons Under tree
canopy Through wet foliage Behind terrain
features Under ground
8
The Importance of Intelligence

• The more the nodes know and understand,
• the less they need to communicate

• If they know and understand everything,
• they dont need to communicate anything

• Putting intelligence and knowledge in the nodes
  
• reduces the need for communication between them

Intelligence is the ability to use knowledge to
make decisions and act appropriately in an
uncertain and dynamic environment
9
Basic Intelligent System
Perception establishes correspondence between
internal world model and external real world
Behavior uses world model to generate action to
achieve goals
10
Intelligent System Architecture
4D/RCS Reference Model
6
11
Intelligent System Architecture
4D/RCS Reference Model
6
12
A 4D/RCS Computational Node
4D/RCS node
13
4D/RCS Reference Model Architecture for Unmanned
Vehicle Systems
Developed for U.S. Army Research Lab Demo III
program
Adopted by GDRS for FCS Autonomous Navigation
System Adopted by TARDEC for Vetronics Technology
Integration
Hierarchical structure of goals and
commands Representation of the world at
many levels Planning, replanning, and
reacting at many levels Integration of many
sensors stereo CCD FLIR, LADAR,
radar, inertial, acoustic, GPS, internal
14
4D/RCS for FCS
15
Representation of Knowledge
4D/RCS node
SENSE
MODEL
ACT
4
16
Forms of Representation

• Iconic
• - signals, images, maps (arrays)
• - Support geometry, and navigation
• - Have range and resolution in space and time
• Symbolic
• - objects, events, classes (abstract data
  structures)
• - Support mathematics, logic, and linguistics
• - Have vocabulary and ontology
• Links
• - relationships (pointers)
• Support syntax, grammar, and semantics
• Have direction and type

17
Types of Knowledge
About the environment places, conditions,
situations About things entities, states,
attributes, classes, relationships About
actions tasks, skills, motives, plans,
behaviors About experiences events, situations,
scenarios, seeing, hearing, feeling,
temperature, touch About rules logic,
mathematics, language, physics About models
geometry, dynamics, kinematics, simulation,
visualization
18
Value Judgment
Rationale for decision-making and planning
19
Sensory Processing a set of processes
that interpret current observations in the
context of stored knowledge
Focus attention Sensor fusion (LADAR,
color, FLIR, radar, inertial, GPS) Segmentation
grouping Computation of group attributes Classific
ation recognition
20
Segmentation Grouping Pixels to Entities
Iconic to Symbolic
21
LADAR image
22
LADAR Intensity Image Red bright, Blue
dark Focus Camera on On-Coming Cross Traffic
23
Compute Road Edges
24
Segment Road Objects Above It
25
Segment On-Coming Cars from Road Surface
26
Compute Attributes of Segmented Cars
Object2 Range 62 m Closing speed 2
m/sec Width 162 cm Height 140 cm
Object1 Range 41 m Closing speed 2
m/sec Width 176 cm Height 128 cm
27
Attributes of 4D/RCS

• Combines AI with control theory
• Hierarchical representation of tasks, space,
  time
• Combines deliberative with reactive at many
  levels
• Depends strongly on sensing and perception
• Supports a rich dynamic world model at many
  levels
• Integrates prior knowledge with current
  observations
• Models functional architecture of the human brain
• Addresses the full range of human behavior
• Is mature with engineering tools and software
  libraries

28
4D/RCS Documentation
Numerous journal articles, reports, and
conference papers Extensive software library
http//www.isd.mel.nist.gov/projects/rcslib
29
Most Recent Publication 2007
30
Summary and Conclusions

• On the battlefield, communications are
• limited and unreliable

• The more knowledge and intelligence
• systems have, the less they need to
• communicate

• 4D/RCS is a network of intelligent nodes
• configured to be a cognitive system