A System Layout for Cognitive Service Robots Stefan Schiffer1,2 Alexander Ferrein2 1 2 Knowledge-Based Mobile Autonomous Systems and Systems Group (KBSG) Cognitive Robotics Institute (MASCOR) RWTH Aachen University FH Aachen University of Applied Sciences Abstract—In this paper we discuss a system layout for cognitive service robots. The goal is to sketch components and their inter- Na High-level Reasoning tu Int ral L play needed for cognitive robotics as introduced by Ray Reiter. erp an ret gua We are particularly interested in applications in domestic service ati ge on robotics where we focus on integrating qualitative reasoning and Qualitative Spatial Self-Maintenance Representations human-robot interaction. The overall objective is to build and and Reasoning Fuzzy Control Annotations Qualitative Represen- tations & Semantic maintain a knowledge-based system and agent specification. Notions Human I. I NTRODUCTION ot ob s In this work, we are concerned with a system layout for a n-R dule m Mo Hu what is often called cognitive robotics. Cognitive robotics as Base Components sic tion Ba erac Int introduced by the late Ray Reiter is to be understood as “the study of the knowledge representation and reasoning problems faced by an autonomous robot (or agent) in a dynamic and Fig. 1: A cognitive service robot system layout incompletely known world” [5]. Our application domain is domestic service robotics [15]. It deals with socially assistive robots that perform helpful tasks for humans in and around enduring autonomy. It is an extension of the high-level control the house. These robots must be able to engage in communi- that has tight connections to the basic components. cation with the humans around them. What is more, when a A. Basic Human-Robot Interaction Modules robot needs to assist humans with complex and cognitively challenging tasks, it must be endowed with some form of Our domestic service robot is supposed to interact with reasoning that allows to take decisions on the course of action laymen. Hence, it needs to be operable by such laymen and in complex scenarios. In addition, autonomous operation for the interaction between the human and the robot needs to be extended periods of time is only possible if the robot can as natural and intuitive as possible. This is why we argue for handle certain variations and unavoidable errors by itself. Also, extending the basic capabilities with modules for three im- it should be flexible in dealing with human fallibility. We refer portant human-robot interaction components, namely speech, to such a robot as a cognitive service robot system. face, and gesture recognition. Examplary solutions for such components tailored for the particular application scenarios II. A C OGNITIVE S ERVICE ROBOT S YSTEM L AYOUT can be found in [3], [1], and [9] respectively. We consider these components since they represent (perhaps the most) We now discuss a system layout for such a cognitive service important modalities in human-robot interaction. Human-robot robot in domestic applications. Figure 1 shows an overview interaction can be made even more natural and affective with of the elements that we think are necessary and useful for additional components such as text-to-speech and an animated a cognitive robotic system. The particular focus here is on visual appearance. integrating qualitative reasoning and human-robot interaction [7], [8] for applications in domestic domains [11]. B. High-level Reasoning The blue elements are components that provide basic ca- A domestic service robot that needs to assist humans with pabilities like collision avoidance and localization. The green complex and cognitively challenging tasks, must be endowed boxes represent high-level components, that is, components with some form of reasoning that allows it to take decisions featuring a sophisticated reasoning mechanism. We use a in such complex scenarios. This high-level reasoning abstracts logic-based high-level language called ReadyLog [4] which, from the details of lower levels and provides mechanisms among other things, features decision-theoretic planning in to come up with a dedicated course of action for a robot the spirit of [2]. The orange components bridge between to reach a particular goal. Our robot features a logic-based the high-level and the human or extend the high-level with high-level reasoning component for that purpose. It allows for mechanisms to facilitate intuitive interaction. The yellow box flexibly combining programming and planning in the behavior finally, is an optional but desirable component to enable specification of the robot. Proceedings of EUCognition 2016 - "Cognitive Robot Architectures" - CEUR-WS 44 C. Qualitative Representations and Control robotics. The system layout features components that allow One of the issues in developing a robotic system that for implementing a capable service robotic system. The layout interacts with humans is the difference in representations addresses bridging the gap between the robot and the human with humans and machines. A technical system mostly uses with several measures, making available the qualitative notions numbers to represent things like speed, distance, and ori- that humans commonly use in the robot system, in general, entation while humans use imprecise linguistic notions. A and in the high-level reasoning, in particular. This allows robotic system that assists humans in their daily life, must for natural interaction and with its advanced reasoning the be equipped with means to understand and to communicate robot can assist its human users with complex and cognitively with humans in terms and with notions that are natural to challenging tasks. This is especially useful with disabled or humans. The qualitative representations and reasoning with elderly people. them should be available especially for positional information R EFERENCES (e.g. as proposed in [12]) since these are very frequent in [1] Vaishak Belle, Thomas Deselaers, and Stefan Schiffer. Randomized trees domestic settings, for example, with references to objects and for real-time one-step face detection and recognition. In Proc. Int’l Conf. places. on Pattern Recognition (ICPR’08), pages 1–4. IEEE Computer Society, December 8-11 2008. D. Semantic Annotations [2] Craig Boutilier, Ray Reiter, Mikhail Soutchanski, and Sebastian Thrun. Decision-theoretic, high-level agent programming in the situation cal- Another building block to mediate between the raw sensor culus. In Proc. Nat’l Conf. on Artificial Intelligence (AAAI-00), pages data and the numerical information that the base components 355–362, Menlo Park, CA, July 30– 3 2000. AAAI Press. [3] Masrur Doostdar, Stefan Schiffer, and Gerhard Lakemeyer. Robust of a cognitive robot work with are semantic annotations. In our speech recognition for service robotics applications. In Proc. Int’l cognitive robot system, for instance, we allow for generating RoboCup Symposium (RoboCup 2008), volume 5399 of LNCS, pages semantically annotated maps [10]. This attaches semantic 1–12. Springer, July 14-18 2008. [4] Alexander Ferrein and Gerhard Lakemeyer. Logic-based robot con- information to places like functions of a room or where it trol in highly dynamic domains. Robotics and Autonomous Systems, is likely to find people in an apartment. Another example 56(11):980–991, 2008. could be part of the object recognition [6], where objects are [5] Hector J. Levesque and Gerhard Lakemeyer. Cognitive Robotics. In Frank van Harmelen, Vladimir Lifschitz, and Bruce Porter, editors, described by a set of (semantic) attributes. This way, one can Handbook of Knowledge Representation, chapter 23, pages 869–886. dynamically build classes of objects, for example, all objects Elsevier, 2008. with a specific color. [6] Tim Niemueller, Stefan Schiffer, Gerhard Lakemeyer, and Safoura Rezapour-Lakani. Life-long learning perception using cloud database E. Natural Language Interpretation technology. In Proc. IROS Workshop on Cloud Robotics, 2013. [7] Stefan Schiffer. Integrating Qualitative Reasoning and Human-Robot Humans tend to be imprecise and imperfect in their natural Interaction for Domestic Service Robots. Dissertation, RWTH Aachen University, Department of Computer Science, Feb 2015. spoken language. Therefore, when natural language is used to [8] Stefan Schiffer. Integrating qualitative reasoning and human-robot give instructions to a robot, the robot is potentially confronted interaction in domestic service robotics. KI - Künstliche Intelligenz, with incomplete, ambiguous, or even incorrect commands. 30(3):257–265, 2016. [9] Stefan Schiffer, Tobias Baumgartner, and Gerhard Lakemeyer. A Aiming for a robust and flexible system a method for natural modular approach to gesture recognition for interaction with a domestic language interpretation that can account for handling such service robot. In Intelligent Robotics and Applications, pages 348–357. fallibility is beneficial. We present such a system [13] that Springer, 2011. [10] Stefan Schiffer, Alexander Ferrein, and Gerhard Lakemeyer. Football is uses decision-theoretic planning in the spirit of DT-Golog [2] coming home. In Proc. 2006 Int’l Symp. on Practical Cognitive Agents to interpret the instruction given to the robot. It is able to and Robots (PCAR’06), pages 39–50, New York, NY, USA, November account for imprecise and missing information and initiates 27-28 2006. ACM. [11] Stefan Schiffer, Alexander Ferrein, and Gerhard Lakemeyer. C AESAR steps for clarification accordingly. – An Intelligent Domestic Service Robot. Journal of Intelligent Service Robotics, pages 1–15, 2012. F. Self-Maintenance [12] Stefan Schiffer, Alexander Ferrein, and Gerhard Lakemeyer. Reasoning A robotic system that is capable of planning and executing with Qualitative Positional Information for Domestic Domains in the Situation Calculus. Journal of Intelligent and Robotic Systems, 66(1– complex tasks is a complex system itself. That is why such 2):273–300, 2012. a system is itself vulnerable to errors. These errors are not [13] Stefan Schiffer, Niklas Hoppe, and Gerhard Lakemeyer. Natural lan- restricted to action execution but span to internal system errors guage interpretation for an interactive service robot in domestic domains. In Agents and Artificial Intelligence, volume 358, pages 39–53. Springer, as well. As an additional component in the system layout 2013. we proposed a system for self-maintenance [14] that is able [14] Stefan Schiffer, Andreas Wortmann, and Gerhard Lakemeyer. Self- to detect and circumvent certain errors. Thus we increase Maintenance for Autonomous Robots controlled by ReadyLog. In Proc. IARP Workshop on Technical Challenges for Dependable Robots in the system’s robustness and enable longer-term autonomous Human Environments, pages 101–107, Toulouse, France, June 16-17 operation. 2010. [15] Thomas Wisspeintner, Tijn van der Zant, Luca Iocchi, and Stefan III. C ONCLUSION Schiffer. RoboCup@Home: Scientific Competition and Benchmarking for Domestic Service Robots. Interaction Studies, 10(3):392–426, 2009. In this paper, we discussed the layout of a cognitive ser- vice robotic system that integrates qualitative reasoning and human-robot interaction for applications in domestic service Proceedings of EUCognition 2016 - "Cognitive Robot Architectures" - CEUR-WS 45