Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Swarm Intelligence and Swarm AI

Doug Stephen
February 11, 2013

Swarm Intelligence and Swarm AI

A brief introduction to Swarm AI and how it intends to present a general framework for developing Swarm Intelligence.

--- Downloads ---
Keynote, PPT, & PDF (.zip archive): http://d.pr/f/n5Af

Doug Stephen

February 11, 2013
Tweet

More Decks by Doug Stephen

Other Decks in Programming

Transcript

  1. Quantifying the emergent behavior of large, social groups Just look

    at all the birds. Being social. In a large group. Monday, February 11, 13
  2. In Geek Speak ➡ Swarms are made of Agents ➡

    Swarms are made of Intelligent Agents Monday, February 11, 13
  3. In Geek Speak ➡ Swarms are made of Agents ➡

    Swarms are made of Intelligent Agents ➡ Swarms are made of lots of Intelligent Agents Monday, February 11, 13
  4. So what about agents in a swarm? How do they

    solve problems? Monday, February 11, 13
  5. Rule 1 Agents are autonomous, and work individually on a

    common problem. Monday, February 11, 13
  6. Autonomy Agents cannot directly control each other. This means no

    delegation of tasks or jobs. Monday, February 11, 13
  7. Rule 2: Agents are simple. Like this slide. More than

    simple, they should be fast and have limited knowledge. Monday, February 11, 13
  8. No Big Picture Agents don’t need to sense/perceive the whole

    environment; just what’s important to their slice. This keeps them simple and fast, which really means they process their job quickly. Monday, February 11, 13
  9. Rule 3 Agents can communicate indirectly to in uence the

    behavior of other agents. Monday, February 11, 13
  10. Two Ways In uence via stigmergy In uence via observable

    change in state Monday, February 11, 13
  11. Baaah. These principles lead to Emergent Behavior. The swarm resolves

    the big picture problem, without ever having knowledge of it. Monday, February 11, 13
  12. Design of Swarm AI systems doesn’t require a wealth of

    interdisciplinary knowledge. Monday, February 11, 13
  13. The agents are simple to design, but the problem partitioning

    and task delegation might not be. Monday, February 11, 13