Simulating Plausible Movement-based Interactions between Agents and Avatars using Biomechanical Principles and Psychological Factors

Many Virtual Reality-based applications enable the user to actively participate in the virtual environment (VE) by embodying a virtual avatar, and these applications rely on the presence of interactive virtual pedestrians, or agents, to enhance the plausibility of the simulation. A key challenge is to generate plausible movements, in terms of navigation paths and full-body animation, for each virtual agent as it interacts with the user's avatar and other agents. The overall effectiveness of the application depends on both, the realism of the agent's movements, and the faithful representation of the user's avatar. The primary goal of this dissertation is to present novel techniques to simulate plausible movements and behaviors for virtual agents, to synthesize personalized avatars that reflect the user's unique motion, and to generate plausible avatar-agent interactions in immersive VE's. First, we present a novel approach that generates trajectories for agents that reflect the speed and density relationship observed in human crowds. Our approach models the biomechanical relationship between stride length and walking speed, as well as psychological factors such as preference for personal space. Second, we propose a velocity computation algorithm that takes into account human motion constraints and generates plausible full body motion for the virtual agents and movement-based avatar-agent interactions. Third, we present an algorithm that simulates plausible gaze-based behaviors in avatar-agent interactions. Our approach relies on a Bayesian interpretation of the Theory of Mind concept to infer the user's intentions and generates an appropriate response for the virtual agents. Finally, we study the perception of motion rendered on virtual avatars and propose a novel data-driven approach to rapidly synthesize a photo-realistic virtual avatar that reflects the user's unique gait. We highlight the interactive performance of our algorithms in complex virtual environments with tens of virtual agents. We also conduct extensive user studies and show that the interactions generated using our proposed algorithms are perceived as more plausible, and that they have a greater impact on the user in immersive settings.Doctor of Philosoph