Walk-in NPC Behavior Based on RVO Using Waypoints

Behavior is one of the major points for Artificial Intelligence (AI). One of the most common researches regarding AI behavior is the behavior in the navigation system, in order to generate navigation for AI to make it more natural. In some simulations, the pedestrian’s behavior in navigation is not natural, predictable, and causes other unexpected events such as deadlock, and so on. This study employs the navigation system with Waypoints Navigation to determine paths or branches to take, also to restrict and to liberate AI in walking along the path, and the obstacle avoidance system from Reciprocal Velocity Obstacle (RVO) to assist in avoiding obstacles or other agents. Experiment results show that agents have various and unpredicted results in their behavior where first, the adjustment of parameters enables the number of the pedestrians randomly entering the branch path. Second, the possibility of the deadlock occurs can be reduced by 50 % or more, depends on the numbers of agents and value used in the parameters. Third, a balance between the number of pedestrians walking and standing still or idling can be achieved.