Computer Animation and Virtual Worlds

With the advent of multi-core architectures, new approaches to accelerate pathfinding operations are increasingly important. Current parallel versions of such algorithms have a large overhead, yield far from optimal paths, do not scale up to many cores, or are cache unfriendly. Parallel Ripple Search constructs a path by first flooding the search space between start node S and goal node G at ‘equidistant’ positions; as soon as enough ‘ripple collisions’ have occurred, they are used to construct the full path.

In this paper, we propose an online path planning algorithm in dynamically changing environments with unknown evolution such as physically based-environments. Our method represents objects not only in terms of obstacles but also in terms of navigable surfaces. It allows finding temporal paths through disconnected and moving platforms without a priori knowledge on environment evolution.

This paper extends the ‘Parametric Control of Captured Mesh Sequences for Real-Time Animation’ published in the Proceedings of the 4th International Conference on Motion In Games, following the invitation from the conference chairs to submit our work to the MIG2011 Special Issue of the Journal of Computer Animation and Virtual Worlds.

The approach for real-time non-linear mesh blending introduced in the MIG2011 paper is extended in this paper, presenting an evaluation of the method, including the following points:

• Quantitative and qualitative evaluation of the results and comparison with respect to the non-linear approach.
• Analysis of memory requirements for the presented approach in different scenarios.
• Pre-computation time and online time requirements for the presented approach.
• Influence of the threshold epsilon on the final performance.
• New figures to support the discussion: Figure 2 shows an example of an aligned dataset; Figure 4 shows the influence of the threshold to the final result, highlighting the corrected areas; Figure 5 characterises the percentage of corrected area and the maximum displacement error for the presented approach, depending on the threshold used; Figure 6 shows how error and CPU time vary with respect to both the number of subdivisions and the threshold used; and Figure 8 shows two parametric animations interactively created by the presented approach.

The results and evaluation presented in this paper show that our approach provides real-time control of mesh sequences, combining the realism of non-linear blending methods with the real-time performance of the linear approach, whilst maintaining the natural surface dynamics of captured movement.

This paper presents a method that enables twisting, tearing by stretching, tearing by twisting and flicking effects in traditional string simulation methods that explicitly represent a string by particles and segments. Wide variety of strings in our daily lives can be generated by this method. This paper also introduces an optimized grid-based collision detection for accelerating the computation.

The proposed algorithm generates a rigged 3D model by locating the appropriate template skeleton on the extracted curve skeleton without any user's intervention. It can correctly locate all joints inside the input 3D model and also label each joint with the name that is corresponding to motion data. Then, motion data can be transferred to drive an animation automatically. In contrast to previous work, this method is not restricted to standard pose and orientation.