AIRC Team

My research focusses on data-driven animation generation for engaging virtual characters for use in serious games. My supervisors are Dr. Cathy Ennis and Dr. Robert Ross. The focus of my work has been on proposing a novel approach for data-driven character animation capable of robustly generating dynamic, varied, human-like behaviours portraying multiple activities like waving and pointing as well as complex conversation beat gestures. The research contributions from my PhD work involved the use of data-driven reinforcement learning for learning latent dynamics models animation agents, and realism regularisation for agents using adversarial techniques to further enhance the human-likeness of behaviours portrayed. The research lead to the creation of RLAnimate, a framework for human-like character animation generation that can be used by game developers to add human-like character behaviours to games and applications, and researchers to explore reinforcement learning or data-driven animation to explore augmenting the range and complexity of behaviours that can be portrayed, such as complex aspects of behaviour expression like emotion and personality. (If you’d like to find out more about this research, please visit https://virtualcharacters.github.io)