Computer Animation and Virtual Worlds

Games and simulations frequently model scenarios where obstacles move, appear, and disappear in an environment. We show how to maintain a 2D or 2.5D navigation mesh in an environment that contains such dynamic polygonal obstacles. Experiments show that local updates are fast enough to permit real-time updates of the navigation mesh.

This paper presents a novel method to plan interactive task based on Q-learning for intelligent characters. Firstly, in the data preprocessing phase, we abstract the motion clips as high-level behaviors and construct the interactive behavior graph to define the interactive capabilities in terms of interactive features. Secondly, for the controller training phase, Q-learning algorithm is employed to train the controller. Finally, in the motion-synthesis phase, the optimal motion sequences can be generated by following the controller to accomplish the interactive task. The experiments demonstrate that our framework can generate realistic motion sequences to plan interactive task in complex environment.

Using an animation saliency map computed in the high-dimensional space of the character's geometric model and a lowdimensional embedding of the animation sequence obtained using LLE, we extract an optimized keyframe representation of the animation that can reconstruct the original animation with less-perceived error. The keyframe search is carried out in the low-dimensional embedding space, making it efficient. This approach yields much better results than earlier methods for keyframe extraction.

Controlling a virtual avatar within a densely populated environment can be challenging because the surrounding virtual characters may obstruct the user's path. Predicting the user's intentions and steering surrounding agents to avoid the user's future path improve the user's navigability in a crowded virtual environment. Observers report that the “polite” virtual characters are more natural and human-like.