Autonomous agents

Autonomous agents are 3D entities in a virtual environment that perceive, decide, and act on their own without continuous direct human control or purely scripted sequences. They combine sensors (virtual perception of the world and other objects), an internal state or goals, and a decision-making mechanism (such as behavior trees, state machines, or learning algorithms) to determine their actions. In 3D contexts they are used for game NPCs, robotics and drone simulations, and crowd or emergent-behavior modeling. Key concerns include pathfinding, collision and local avoidance, reactive vs. planned behaviors, and sometimes adaptive learning for more complex tasks.

Imagine a guard in a 3D game: the guard uses a navmesh for path planning (e.g., A*), has vision and hearing sensors to detect the player, and a behavior tree that governs patrolling, investigating, and chasing. If the player makes a noise the agent switches to an investigate mode; if it sees the player it plans a path and uses local avoidance so it doesn't clip through objects or other agents. In another use case, dozens of autonomous agents can simulate a crowd at a festival, where each individual follows simple rules (keep distance, move toward goals) producing realistic group dynamics.