Skip to main content

Highlight Photo

Perceiving intentions and goals within virtual worlds

Autonomous agents in action

Autonomous agents in action

Six moving virtual agents are shown. Each is represented by a triangles of a different hue (the “tails” of decreasing intensity attached each lead triangle are included to represent the path of the moving agents in this static figure). Agents compete with one another to capture food (circles) while avoiding the obstacles (large squares) and each other. Agents are programmed to choose from among four basic actions: explore, gather food, fight another agent, or flee from another agent. An evolutionary algorithm, which adapts the agents to the environment, selects the most appropriate action at each interval of time. Agents compete against one another, earning points when food is found, captured, and returned to the agent’s “cache” location, while human observers attempt to determine the goals, strategies and intentions of the agents.

Credits: Steve Cholewiak, Peter Pantelis, Paul Ringstad, Kevin Sanik, Ari Weinstein and Chia-chien Wu