Please enable JavaScript.
Coggle requires JavaScript to display documents.
Knowledge-based Systems and Multi-agent Environments - Coggle Diagram
Knowledge-based Systems and Multi-agent Environments
Logical Agents
Humans - not just reflex reactions
Reasoning based on internal representations of knowledge
Pathfinding agents limited
Need a knowledge base
Knowledge Base
Sentence = assertion about the world
Axiom: Given, not derived
Knowledge representation language
TELL: Add Sentence
ASK: Query Knowledge Base
May involve inference
Sequence of Actions
When agent is called
MAKE PERCEPT SENTENCE: Tells knowledge base what it perceives
MAKE ACTION QUERY: Asks knowledge base what action to perform
MAKE ACTION SENTENCE: Tells knowledge base what action was chosen, then executes it
Knowledge Level vs Implementation Level
Declarative approach: Tell it sentences
Eg. Taxi knows route, doesn’t matter how geography of environment is represented
Procedural approach: Encode in code
Wumpus World
What's in the world?
Wumpus
Agent with 1 arrow
Gold treasure
Bottomless pit
P.E.A.S.
Performance Measure
+1000: Agent gets out with gold
-1000: Eaten by Wumpus or falls into pit
-1: Each action
-10: Using arrow
Environment
4x4 grid
Agent at [1,1]
Random gold and Wumpus
0.2 probability of a bottomless pit
Actuators
Forward, turn left, turn right
Grab gold (if on the same tile)
Shoot (arrow goes in direction agent is facing)
Climb out at [1,1]
Sensors
SMELL: Stench from adjacent squares to Wumpus
FEEL: Breeze from adjacent squares to pit
SEE: Glitter from gold
FEEL: Bump from wall
HEAR: Scream from dead Wumpus
Percepts as list: [stench, breeze, none, none, none]
Environment Characteristics
Discrete, static, single agent, sequential, partially observable with unknown model
Multi-agent Environments
Planning: Cooperation and coordination
Each agent makes its own plan
Coordination for shared goal
Conventions
Eg. Tennis doubles - stick to your side of court
Drive on the left
Communication
Plan recognition: Partner moves towards net
Biological examples
Feeding frenzy, honey bee swarm, ants - self-organization, migration, salmon run, bats, zebra stripes
Boids Model by Craig Reynolds
Separation: Avoid crowding (-ve)
Alignment: Head in average direction (+ve)
Cohesion: Steer towards average position (+ve)
Emergent behavior of flying as a pseudo-rigid body
Emergence
Macro-level phenomenon
Patterns of behavior from simple rules
Conway's Game of Life
Rules and Patterns
Overcrowding, loneliness, birth, survival
Examples: Still life (beehive, block, boat), Oscillator (blinker, toad), Gliders
Complexity and Applications
Cellular automata, computer viruses, human diseases, generative music, AI LIFE
ALIFE
Evolution in action
Behavior and intelligence
Synthetic biology
Collective dynamics
Art, music, philosophy of ALIFE
Collective Intelligence
Decision-making in everything from bacteria to humans
Neural networks: Adaptive systems, natural intelligence, not AI (rule-based learning)
Links
http://www.red3d.com/cwr/boids/
http://www.vergenet.net/~conrad/boids/pseudocode.html
http://www.youtube.com/watch?v=eakKfY5aHmY&feature=fvwrel
http://www.sean.co.uk/a/science/conwaysgameoflife.shtm
http://www.emergentuniverse.org/#/life
http://www.physics.buffalo.edu/gonsalves/Java/GameOfLife
lexicon.html
http://www.ted.com/talks/joshua_klein_on_the_intelligence_of
crows.html