Enhancing real-time animation: Ensuring distinctiveness in crowd dynamics through physics-based collision avoidance

Abstract

Crowds are essential in daily activities, performing as dynamic systems of human interaction. They involve numerous individuals collecting in specified locations for diverse activities, whether in urban environments, public events, or social interactions. In virtual environments, animated crowds frequently display repetitive behaviors and a deficiency in movement diversity, leading to unrealistic simulations. It is imperative to provide distinctive and diverse actions for each character, avoiding visual duplication to augment realism. This paper introduces a method to enhance crowd variety by ensuring distinct and realistic character movements. We propose providing various motion types to prevent the duplication of cloned characters. Our approach creates a set of animations and utilizes techniques to control character velocity, ensuring distinctive and convincing movements. Furthermore, we present collision prevention methods based on Newton's Laws, the conservation of momentum, and the laws of kinetic energy. Ray-casting determines collision velocities by considering each character's mass and velocity without external forces. We implement a hybrid pool object approach and occlusion culling techniques to optimize real-time performance, increasing FPS (framed per second) and reducing computational load. These experiments evaluate the efficacy of our technique under different conditions. The results demonstrate the method's effectiveness and flexibility in dynamic environments.