A pratical approach to Humanoid Introspection

A pratical approach to Humanoid Introspection
of 2
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
  A practical approach to Humanoid Introspection Ignazio Infantino 1 , Giovanni Pilato 1 , Riccardo Rizzo 1 and Filippo Vella 1 1 ICAR - CNR, Viale delle Scienze ed.11, Palermo, Abstract We describe an approach for a humanoid robot to understandits internal state (Infantino et al. (2013)). The method isbased on self observation and communication with the ex-ternal world, according to the idea of introspection given bySloman (2010). The robot introspection arises from infor-mation about physical components and software modules.This information is translated in a spatial representation of the hardware and software components of the robot througha SOM, which links the state representation of the robot withanhighlevelrepresentationgivenbyanontology. Theontol-ogy is furthermore linked to a linguistic module that makesit possible the interaction with human beings though a con-versational agent. Introduction The interaction between man and robot can benefit frommodels of the human mind and its cognitive abilities. Oneaspect that has not received enough attention is the mecha-nism that allows the robot to have a kind of self-awareness(Birlo and Tapus, 2011). Self-awareness, including the bodyphysical parts and management functions related to them,is often referred to an a priori knowledge and it is usuallyimplicitly integrated into the architecture. We will illustratea possible approach for the realization of an introspectivecapacity (at present mainly oriented to the physical compo-nents of the robot).Based on an empirical approach, our idea of introspectionof the robot, starts from the analysis of information obtainedautomatically by the embedded software and its related doc-umentation, in particular regarding the relationship (director explicit) between the physical components and softwaremodules. This documentation is used to construct a repre-sentationofthehardwareandsoftwareoftherobotonamap,based on Self Organizing Map (SOM), similar to the hu-man somatosensory map. This map structure can be used toquickly retrieve information semantically related (Honkelaet al., 1997). From this map we can use different approachesto ascend to a higher level of abstraction. In particular, theapproach used in this paper involves a simple association be-tween labels arising from data and ontology entities, tryingto get the expressiveness with a enormous knowledge basedon common sense using Cyc (see The Introspection Architecture System Our approach to introspection is based on self observationand communication. Figure 1 shows the proposed architec-ture. Considering the definition of Sloman, self observationis what the robot should do in order to build, represent andunderstand its internal state. In particular it is necessary tohave a set of sub-systems dedicated to make a snapshot of the Nao robot state. Some systems are supplied by Nao sys-tem software, while some others are developed ad-hoc. Thedataobtainedareusedtobuildarichstaterepresentationthatshould be supplied to an ontology that associates a meaningto the internal state. Our approach integrates static and dy-namic information on robot operation.Static information is related mainly to robot hardwareparts and software modules, like hardware drivers, or mod-ules that supplies some services like face tracking. But theseparts can be active or not during robot operation, and theirstate is part of the robot state itself. A simple list of theseparts can be difficult to manage. In order to obtain a man-ageable state representation we decided to develop a mapthat collects all the robot parts, and to highlight on this mapall the parts that are involved in any robot operation. Such amap can be obtained using the information contained in therobot documentation that is rich of hardware and softwaredetails. Dynamic information is also represented on the mapbut they came from robot operation. When the robot is op-erating it is possible to highlight on the map the units corre-sponding to the active modules and the unit correspondingto the hand. This is a part of the robot state representation.According to figure 1 this Nao State Representation is sup-plied to a Semantic Bridge that analyzes the representationand gives as output a set of information (semantic labels)that are used to activate the right concepts on the ontology.The Linguistic Level exploits these activations in orderto perform a verbal interaction with the human user. In thepresent implementation the Semantic Bridge is constituted Bioinspired Robotics 1005 ECAL 2013 DOI:  Figure 1: The proposed architecture for introspectionby a set of labels. Self Observation The Nao state representation is obtained by using the infor-mation in documentation, the list of processes running onNaosystem, andtheinformationfromsensorinput. TheSelf Observation module builds a representation of the internalstate of the robot, mixing together the information relatedto the robot hardware and the information available aboutsoftware modules running during robot operation. This rep-resentation should be rich enough to represent the percep-tion of the robot body from sensors, to distinguish differentkind of sensors, and also to describe the list of active pro-cesses that control the robot movements and any other usefulstate component.The robot documentation collects informa-tion about the robot hardware and software, so it is a goodstarting point for this kind of representation. In order to ob-tain a suitable document representation, useful for cluster-ing, a sequence of standard pre-processing techniques wereapplied. A SOM(Kohonen, 1995) architecture has been usedto cluster semantically related documents. The sensory in-put receives the visual information from the cameras andfilters the input to select the most promising areas for thedownstream tasks. Nao software platform supplies a list of processes available on the Nao system.These scripts are alsomapped in the document SOM using a suitable set of key-words. Finally a third part of the Nao state is obtained usingsome custom scripts that output the list of processes runningon the Nao system. The Semantic Bridge and Linguistic Level The semantic bridge subsystem obtains a set of suitable la-belsfromtheNaointernalstaterepresentation.Atpresent theimplementation of the semantic bridge uses a look-up tablethat connects the map state images to the labels in the ontol-ogy concepts.The Nao robot has an internal knowledge of its physi-cal structure and functionalities. This knowledge can beexploited to support direct communication on the percep-tion capabilities of the robot, and to describe his state to anhuman interlocutor by using natural language. Moreover,modern semantic tools and introspection capability can beexploited together to improve and support direct communi-cation on the robot perception mechanisms (Infantino et al.,2012; Augello et al., 2013).The Cyc knowledge base (KB) has been used to code re-lations, concepts, constraints, and rules regarding the Naorobot domain. These concepts have been organized in orderto fulfil the self-observation task.The linguistic level is aimed at interpreting natural lan-guages query given by the user. This level exploits a classi-cal pattern-matching technique enhanced with Cyc ontologyinference capability. This feature is obtained by transform-ing natural languages requests into symbolic queries, ex-pressed in the ontology language. Such commands are for-warded to the ontology engine that computes the appropriateinferences and gives results in a symbolic form. The sym-bolic answers are then transformed by the linguistic moduleinto natural language sentences that are finally shown to theuser. The linguistic level has been implemented by using theA.L.I.C.E. web bot (see References Augello, A., Infantino, I., Pilato, G., Rizzo, R.,and Vella, F. (2013). Introducing a creativeprocess on a cognitive architecture.  Biologi-cally Inspired Cognitive Architectures , avilable online at: S2212683X13000467.Birlo, M. and Tapus, A. (2011). The crucial role of robotself-awareness in hri. In  Proc. of HRI 2011 , pages 115– 116. ACM.Honkela, T., Kaski, S., Lagus, K., and Kohonen, T. (1997).Websom - self-organizing maps of document collec-tions. In  Neurocomputing , pages 101–117.Infantino, I., Pilato, G., Rizzo, R., and Vella, F. (2012). IFeel Blue: Robots and Humans Sharing Color Repre-sentation for Emotional Cognitive Interaction. In  Bio-logically Inspired Cognitive Architecures 2012 , volume196, pages 161–166.Infantino, I., Pilato, G., Rizzo, R., and Vella, F. (2013). Apractical approach to humanoid introspection.  Intl. J.of Advanced Robotic Systems. , 10:246.Kohonen, T. (1995).  Self-Organizaing Maps . Springer,Berlin.Sloman, A. (2010). An alternative to working on machineconsciouness.  International Journal of Machine Con-sciousness , 02(01):1–18. Bioinspired Robotics ECAL 2013 1006
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks