IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 53, NO. 5, OCTOBER 2006 1723 Tripodal Schematic Control Architecture for Integration of Multi-Functional Indoor Service Robots Gunhee Kim, Member,IEEE, and Woojin Chung, Member,IEEE Abstract—This paper discusses a control architecture and a sys- tem integration strategy for multifunctional indoor service robot public service robot (PSR) systems. The authors have built three versions of the PSR systems for four target service tasks, which are delivery, patrol, guidance, and floor cleaning. They clarify the requirements of the architecture in their applications, and propose the tripodal schematic control architecture as the solution to the architectural problems. The key idea of proposed architecture is to integrate robot systems using following three frameworks, layered functionality diagram, class diagram, and configuration di- agram. The proposed architecture was successfully evaluated and implemented to PSR platforms for their target tasks. Experimental results clearly showed that the developed strategy was useful for developing the autonomous service robots. Index Terms—Control architecture, Petri nets, robot system integration, service robots. I. I NTRODUCTION T HE public service robot (PSR) systems have advanced toward indoor multifunctional service robots at the Korea Institute of Science and Technology (KIST). Through the field studies of typical public environments such as hospitals, office buildings, and museums, we defined four target tasks: delivery, patrol, guidance, and floor cleaning. From this foundation, we developed three versions of the PSR platforms including office service robots PSR-1 and PSR-2 and a tour guide robot Jinny, as shown in Fig. 1. This paper discusses a control architecture and a system integration strategy for multifunctional indoor service robot PSR systems. Well-defined control architecture is essential in an autonomous robotic system [1], [2]. Control architecture integrates different kinds of hardware and software modules. In some cases, the functional performance of each module is highly dependent on the architecture. Also, the architecture plays important roles in maintenance tasks such as revision of existing components or addition of new modules. Manuscript received July 7, 2005; revised October 28, 2005. Abstract pub- lished on the Internet July 14, 2006. This work was supported by the Ministry of Science and Technology of Korea, performed for the Intelligent Robotics Development Program, one of the 21st Century Frontier R&D Programs. G. Kim was with the Intelligent Robotics Research Center, Korea Institute of Science and Technology, Seoul 136-791, Korea. He is now with the Robotics Institute, Carnegie Mellon University, Pittsburgh, PA 15213 USA (e-mail: gunhee@cs.cmu.edu). W. Chung is with the Department of Mechanical Engineering, Korea Univer- sity, Seoul 136-713, Korea (e-mail: smartrobot@korea.ac.kr). Digital Object Identifier 10.1109/TIE.2006.881956 A. Related Work Due to the importance of the robot architecture, many related research activities have been carried in this area of study. After the conventional sensor-plan-act paradigm, subsumption archi- tecture [3] was introduced. Since then, many control architec- tures have converged to a similar structure based on the hybrid approach which integrates reactive control and deliberation. The historical background and the comparison of architectural paradigms are summarized in [4]–[6]. AuRA [7] is known as the first example of hybrid delib- erative/reactive architecture. In this architecture, the planner constructs specific configurations, which integrate behavioral, perceptual, and environment information. This configuration- based approach has been implemented successfully into sev- eral systems and has inspired the development of many other control architectures. In the early 1990s, the concept of the three-layered hybrid architecture was introduced to robotics by several different groups of researchers in parallel. The typical models of this architecture are SSS [8], 3T [9], and ATLANTIS [10]. They are the architectures that use three levels of abstrac- tion for the coordination of deliberate activities with real-time behaviors in a dynamic environment. Saridis’ intelligent control architecture [11] focuses on the development of analytic models and formal theories for the three-layered architecture of general intelligent machines. Al- though this architecture supports a well-defined mathematical formulation of hierarchical control systems, it does not provide overall architectural solutions of a general mobile robot. For example, it does not consider the particularity of navigation in a dynamic environment. It was only implemented on a PUMA manipulator robot. TCA [12], with no explicit layers, acts like a high-level robot operating system (OS) based on the hybrid approach. For deliberation, TCA constructs a task tree, which is a hierarchical representation of task/subtask relationships; for reactivity, TCA supports several mechanisms for monitoring changes in the environment. Saphira [13] is a robot control system for the Flakey mobile robot project at SRI International. Although the layer is not explicitly illustrated in this architecture, it also adopts the hybrid approach. In the top of the architecture, the PRS-lite instantiates routines for planning, execution monitoring, and perceptual coordination for deliberation; fuzzy behaviors are located at the control level for reactivity. The local perceptual space is located at the center of the architecture so it can provide a coherent geometric representation of the space around the robot. BERRA [14] is also one of typical three-layered hybrid 0278-0046/$20.00 © 2006 IEEE