The Research Of Motion Planning And Path Planning In The Automatic Generation Of Computer Animation

As artificial intelligence, computer graphics and computer software and hardware technology rapidly develops, computer animation has been widely applied in engineering, scientific research, entertainment, and other fields.Full life cycle computer aided animation generation technology, which was proposed by Prof. Ruqian Lu in the 1990s, integrates artificial intelligence and methods based on knowledge into the whole process of animation generation.The technique receives as input a proper story written in some limited natural language and generates a 3D animation cartoon as output.The whole process is handled by computer with the support of knowledge bases.Applying this technology, Swan system and automatic animation generation system based on some powerful 3D software and hardware are successively developed.Motion planning, one of the core modules in animation system, firstly generated qualitative action description(ADL) in qualitative layers, after which qualitative description is transformed into quantitative description(CAL) by action calculation module which also processes specific problems in the calculation procedure.Motion planning is the key of animation generation, for all action data are provided by this module.Path planning, an important subproblem of motion planning in the quantitative layer, can generate a non-collision, real and reasonable path from start point to end point in 3D space by path classification and intelligent method. This paper designs and realizes a motion planning system and a path planning subsystem in the automatic animation generation system.In motion planning, we first focus on the design of system animation intermediate language.Action ADL makes a reference to Swan system ADL standard, and contains action qualitative attributes which are extracted from comprehensive and practical action as key attributes and parameters. Action CAL contains the role action keyframes as unit, and is used for the final animation generating.Then we acuqest BVH format file that recording role actions captured by Motion Capture, and transform the BVH data into action database.At last, we design and realize a complete process of action calculation module in the quantitative layer to transform the action ADL to CAL.The motion information in action database is recorded by Motion Capture System in BVH format.The final 3D animation is rendered in Maya system.However, Maya does not support BVH format but rather uses Euler angles to represent rotations, which lead a cross-platform data conversion problem.This paper presents a transformation algorithm to realize the BVH motion data conversion in Maya.Firstly, for single human skeletal joint, rotation matrices equivalency is used to build correspondences between BVH and Euler rotations, thus Euler angles can be computed.Then the method is applied to the complete skeleton to transform an action.On the path planning problem in 3D spaces, we propose terrain modeling algorithm and obstacles modeling algorithm to respectively partition the scene into subspaces and preserve the subspaces into octree structure, which can present both terrain and object model in a unified form.In view of that the octree is not easy to express the adjacent nodes, we design a neighborhood search algorithms for querying the adjacent nodes.To find a path in the animation scene model, we first introduce a scene graph adjlink structure, which is a dynamic structure to preserve the adjacency relationship of the octree.According to the different animation plots, we put the problem into 3 categories, which are flying path planning, walking path planning and groups path planning.IDA* algorithm is adopted in this paper, and will be adjusted in specific scenario to generate a right path.