Abstract: Group animation is a continuous challenge in computer animation. In this paper, a group animation framework based on multiple autonomous agents. Each animated character in a group is modeled as an autonomous agent, which can perceive information from the virtual environment, generate appropriate intentions, manage rational behaviors, and create motions to realize the intention. Contrary to traditional intelligent animated character techniques, motions for the animated character are generated using a motion capture data, instead of modeling the complex motion generation mechanism. First a fundamental motion library is built using motion capture system, then appropriate fundamental motions are selected from the fundamental motion library, and desired motions are synthesized from these fundamental ones. The animated characters manage their movement autonomically in the virtual environment. To produce the required group animation, animators only need to specify some simple information and record the useful group motion. The experiment shows the effectiveness of this framework.