Truong-Huy Dinh Nguyen

Modeling Warmth and Competence in Virtual Characters

Developing believable virtual characters has been a subject of research in many fields including graphics, animations, artificial intelligence, and human-computer interaction. One challenge towards commoditizing the use of virtual humans is the ability to algorithmically construct characters of different stereotypes. In this paper, we present our efforts in designing virtual characters that can exhibit non-verbal behaviors to reflect varying degrees of warmth and competence, two personality traits shown to underlie social judgments and form stereotypical perception. To embark on developing a computational behavior model that portrays these traits, we adopt an iterative design methodology tuning the design using theory from theatre, animation and psychology, expert reviews, user testing and feedback. Using this process we were able to construct a set of virtual characters that portray variations of warmth and competence through combination of gestures, use of space, and gaze behaviors. In this paper we discuss the design methodology, the resultant system, and initial experiment results showing the promise of the model.

Bootstrapping Monte Carlo tree search with an imperfect heuristic

We consider the problem of using a heuristic policy to improve the value approximation by the Upper Confidence Bound applied in Trees (UCT) algorithm in non-adversarial settings such as planning with large-state space Markov Decision Processes. Current improvements to UCT focus on either changing the action selection formula at the internal nodes or the rollout policy at the leaf nodes of the search tree. In this work, we propose to add an auxiliary arm to each of the internal nodes, and always use the heuristic policy to roll out simulations at the auxiliary arms. The method aims to get fast convergence to optimal values at states where the heuristic policy is optimal, while retaining similar approximation as the original UCT at other states. We show that bootstrapping with the proposed method in the new algorithm, UCT-Aux, performs better compared to the original UCT algorithm and its variants in two benchmark experiment settings. We also examine conditions under which UCT-Aux works well.