Evolving collective behavior in an artificial ecology

Collective behavior refers to coordinated group motion, common to many anim als. The dynamics of a group can be seen as a distributed model, each "anim al" applying the same rule set. This study investigates the use of evolved sensory controllers to produce schooling behavior. A set of artificial crea tures "live" in an artificial world with hazards and food. Each creature ha s a simple artificial neural network brain that controls movement in differ ent situations. A chromosome encodes the network structure and weights, whi ch may be combined using artificial evolution with another chromosome, if a creature should choose to mate. Prey and predators coevolve without an exp licit fitness function for schooling to produce sophisticated, nondetermini stic, behavior. The work highlights the role of species' physiology in unde rstanding behavior and the role of the environment in encouraging the devel opment of sensory systems.