Ubiquitous Korea Project

ion in general can be classified into control abstraction and data abstraction. Control abstraction is the abstraction of actions while data abstraction is that of information or knowledge. The result of control abstraction is often represented as a framework while data abstraction as an ontological data model. In this article, we present an agent-based control framework based on the traditional three-layer robot architecture [10] to provide appropriate level of control abstraction enabling integration of knowledge and functionalities for intelligent service robots. The framework is designed borrowing the core concept of the serviceoriented modeling [3] and model-driven architecture [14] to provide data abstraction in terms of ontological definition of shared knowledge. In the following sections, we first motivate our approach to control and data abstraction for an intelligent service robot and present the framework with design rationales with experimental results. 3.2 Task and Knowledge Management Framework Comprehensive range of diverse functionalities and unavoidable complexity in developing intelligent service robots call for a unified but simple control of the functional components. The diverse functionalities include recognition, speech understanding, manipulation, expression, navigation, reasoning, decision making, planning, and so on. The most important lesson learned from the past five years of development of intelligent service robots is the significance of the framework to handle the complexity of integration. Our framework is designed based on the following observations about an intelligent service robots. • Intelligent system: An intelligent service robot is by definition is an intelligent system. According to Erman [7], intelligent systems (1) pursue goals which vary over time, (2) incorporate, use, and maintain knowledge, (3) exploit diverse, ad hoc subsystems embodying a variety of selected methods, (4) interact intelligibly with users and other systems, and (5) need to be introspective and aware of their progress in applying their knowledge and subsystems in pursuit of their goals. • Heterogeneous system: An intelligent service robot consists of multiple functional components that are heterogeneous in terms of implementation languages and platforms. • Distributed system: The components and information to be integrated as an intelligent service robot may be distributed over multiple, typically single board computers, to satisfy the required processing power and timeliness and to properly handle parallel and layered activities. • Evolving system: The performance goals for an intelligent robot system are continually increasing in complexity. 1264 Minkoo Kim et al. Fig. 3 Task and Knowledge Manager(TKM) in the deliberative layer of the underlying three-layer robot architecture consist of two parts: The Task Manager(TM) provides interface to the Control Components in the sequencing layer for a consistent and unified control view to the functional components. Knowledge Manager(KM), on the other hand, provides data interface as an information integration framework so that the distributed and heterogeneous components may share information and interact consistently.