The research on motion planning of a virtual agent is very important in computer animation, where path planning is the most representative. In this paper we do an in-depth study about the problem of static and global path planning for three-dimensional virtual animals in the marine environment, constructed the path planning strategy of a fast randomized algorithm. First of all, we carry out a pre-processing on the environment, proposed the new partial random sampling strategy and the sequential connecting strategy, which to generate "roadmap" of three-dimensional free space. And then, we put forward a driven step by step inquiring strategy, analyzed two feature-parameters and time complexity of the inquiring strategy. Finally, the simulation experiments show the effectiveness of algorithm in solving the issue of static and global path planning in three-dimensional.
