在多人虛擬環境系統的設計上，讓使用者能自行設計化身的行為，並即時以動畫元件的形式安裝到虛擬世界中，是3D內容能否達到共享的關鍵。本論文主要研究的部分包含三個部分：第一個部分是提供虛擬環境系統動態載入動畫元件的機制，使得虛擬環境系統可以動態加入客製化之化身行為；第二個部分是在現有的虛擬環境系統中加入語意的描述，並增加互動訊息傳送時的彈性；第三個部分是實現可客製化的動畫元件，分別以化身和環境間以及化身彼此間的互動，來說明上述機制的可行性。在動態的載入動畫元件的部分，客製化的動畫元件在系統上得以即時安裝並執行。動畫元件在安裝上可使用XML片段，並交由OSGi Framework中的服務來處理此XML標籤。另外，在加入物件的語意描述後，使得這些動畫元件可以取得世界的資訊，並進一步產生符合當時環境限制或應用需求的動畫。我們以Ontology來描述環境和化身的資訊，並實際製作路徑規劃器元件和化身間互動元件兩個範例。我們利用動態安裝及語意資訊兩個機制，以實例說明如何達到實現化身可客製化行為之目的。 / In the design of multi-user virtual environments, in order to share 3D contents designed by users, it is crucial to allow the behaviors of an avatar to be designed as animation components and loaded at the run time. In this thesis, we attempt to address the problem of designing a semantic virtual environment by considering the following three parts. First, we have designed a mechanism for user-designed animation procedures to be installed and loaded at run time. Second, we have augmented our virtual environment system with semantic descriptions and enhanced the flexibility of message interchange. Third, we have used two types of interaction scenarios, avatar-environment and avatar-avatar, to illustrate how customized animation components can be designed to enhance the functionality of a virtual environment. In our system, we allow users to design their own XML tags and the corresponding animation components managed in the OSGi framework. These components can acquire world information and generate appropriate animations according to application requirements and environment constraints. We have used ontology to describe the semantics of environments and avatars. Two example components: the motion planner and the avatar-avatar interaction have been designed to illustrate the dynamic installation process and the retrieval of semantic information for the realization of customizing avatar behaviors.