Swarm Intelligence

Winter 2004 Lab 03/04 Lecture Supplement

Apr 5, 2004 Intelligence

• New field of study (<10 years old) • Study of models based on natural models (usually of social insects) • Part of Self-Organization that studies emergent phenomena in systems of interacting agents Components

• Agents: – Interact with the world and with each other (either directly or indirectly) – Simple behaviours – e.g. ants, termites, bees, wasps • Communication: – How agents interact with each other – e.g. pheromones of ants Main Idea

• Simple behaviours of individual agents + Communication between a group of agents = Emergent complex behaviour of the group of agents • i.e. global behaviours arising from interactions between agents using simple local behaviours without any global controller • This is an example of self-organization Communication

•Direct: • e.g. Touching, vision, smelling other agents • Pretty obvious interactions • Indirect: • : – One agent modifies the environment – Another agent reacts to the modification of the environment – e.g. ants leaving pheromones in the environment (when returning to the nest with food) which are picked up by other ants and followed to the food source – e.g. termites leaving soil pellets on the ground and triggering different placements based on the size of the pellet structure, resulting in nest construction Algorithms

• Using agents and usually stigmergic behaviours • Agent interactions are modeled by a set of action -> reaction rules • e.g. if concentration of pheromones is low, I wander around; if concentration is high enough, I go towards the highest concentration • The rules are applied locally – the agents do not have global information, only know about their local environment • Through this local rule application, global emergent behaviours are observed Applications

• Optimization problems • Computer approximate solutions to problems such as combinatorial problems (e.g. TSP) and routing (e.g. network routing) • Construction algorithms • Using self-organization and stigmergy to build structures using a set of simple rules (based on wasp/termite nest building) • Self assembling robots • Cooperative transport • Accomplishing tasks (in this case transport) that require multiple agents cooperating with each other and without global information • Robotic cooperative transport Boids

• Model by Craig Reynolds • http://www.red3d.com/cwr/boids/ • Model of life-like artificial of agents (called boids) using 3 simple rules • Demonstrating interesting global flocking behaviour from simple local rules • Each boid can see other boids inside its sphere of neighbourhood • Can be easily adapted to demonstrate other behaviours such as fish schooling Boid Rules

• Separation: • Avoid crowding local agents • Alignment: • Steer towards average heading of local agents • Cohesion: • Steer to move towards average position of local agents Good Books

: From Natural to Artificial Systems by Bonabeau, Dorigo, Theraulaz (~$40) • http://www.amazon.ca/exec/obidos/ASIN/0195131592

• Self-Organization in Biological Systems by Camzine, Deneubourg, Franks, Sneyd, Theraulaz, Bonabeau. (~$50) • http://www.amazon.ca/exec/obidos/ASIN/0691116245