Anchoring symbols to conceptual spaces: the case of dynamic scenarios
Introduction
A cognitive architecture for robot vision has been proposed by the authors in [4], [5], [6]; it is aimed at the representation of knowledge extracted from visual data in both static and dynamic scenarios. One of the main assumptions underlying the design of this architecture is the need of a principled integration of the approaches developed within the artificial vision community and the symbolic, propositional systems developed within symbolic knowledge representation (KR) in AI. Such an integration is based on the introduction of a conceptual level of representation, intermediate between the processing of visual data and declarative, propositional representations.
This paper deals with the anchoring of one of the most influential symbolic formalisms adopted in cognitive robotics, namely the situation calculus, to the conceptual representation of dynamic scenes. We discuss in particular how actions, situations and fluents may be anchored (in the sense of anchoring developed by Coradeschi and Saffiotti [9], [11]) to the representations at the conceptual level, which are in turns generated starting from the robot perceptions (for an up to date survey on different perspectives on anchoring see [10]).
The main motivation for choosing the situation calculus lies in the fact that it is one of the simplest, more powerful and best known logic formalisms for the representation of knowledge about actions and change. It was primarily developed by McCarthy and Hayes [22]; for up to date and exhaustive introductions see [28], [27]. Nowadays, it is a widely adopted formalism in the cognitive robotics literature; efficient Prolog implementations have been proposed [12], [13], [21]; simplified versions of the situation calculus are used by working mobile robots [3], [14], [18].
The following discussion is based on an experimental setup aimed at obtaining an intelligent visual control of a robotic finger starting from visual data. The finger has been entirely developed at the Robotics Laboratory, Department of Computer Engineering, University of Palermo. It is made up by three phalanxes: a terminal phalanx a, a middle phalanx b and an upper phalanx c (see Fig. 1).
The finger is driven by schematic behaviors [1], and performs articulated movements, such as pushing a ball (Fig. 2) or picking up torus-shaped objects (Fig. 3). The system is equipped with a video camera that acquires the movements of the objects and of the finger itself, in order to perform intelligent monitoring operations. The acquired visual data are anchored to symbolic descriptions of the finger operations.
The system takes in input a sequence of images corresponding to subsequent phases of the evolution of the scene (the movements of the robotic finger and their effects on the whole scene), and produces in output a declarative description of the scene, formulated as a set of assertions written in the formalism of the situation calculus.
Such a symbolic description may be employed to perform high-level inferences, e.g. those needed to generate complex long-range plans, or to perform causal and diagnostic reasoning about the system operations. Symbolic assertions may also be used to generate explanations of the operations of the finger, in order to perform high-level teleautonomy [8].
The paper is organized as follows. In the next section, the main assumptions underlying the cognitive architecture are summarized. The third section is devoted to a synthetic description of the conceptual level representation of motion. The fourth section shows in details how the situation calculus is anchored to the conceptual representation. The last section discusses the proposed framework, and compares it to some relevant frameworks for anchoring described in the literature. Short conclusions follow.
Section snippets
The cognitive architecture for visual perception: an overall view
The existing attempts to integrate visual perception with propositional KR are mostly oriented towards natural language interpretation, with particular emphasis on man–machine interaction. They face only in a marginal way the general aspects of knowledge representation (see [6] for a review).
Our proposal is based on the hypothesis that a principled integration of the approaches of artificial vision and of symbolic KR requires the introduction of an intermediate representation between these two
Conceptual spaces for representing motion
As previously stated, representations in the conceptual area are couched in terms of a conceptual space [15] that provides a principled way for relating high level, linguistic formalisms on the one hand, with low level, unstructured representation of data on the other. In this sense, we claim that conceptual spaces are a valuable tool for anchoring [7]. A conceptual space CS is a metric space whose dimensions are in some way related to the quantities processed in the subconceptual area.
Anchoring situation calculus to conceptual spaces
In the linguistic area, the evolution of the conceptual space is represented in terms of logic assertions expressed in the situation calculus formalism. Indeed, the representation adopted by the situation calculus is in many respects homogeneous to the conceptual representation described in the previous section.
In order to anchor linguistic area expressions to structures in the conceptual space, an anchoring function Φ associates expressions of the situation calculus to their counterpart in the
Discussion
In the last few years, the problem of anchoring symbols to data coming out of sensors became a relevant topic in autonomous robotics, and several proposals have been developed. In particular, a model which presents similarities with our approach is due to Coradeschi and Saffiotti. Briefly, our linguistic area corresponds to their Symbol system, and our subsymbolic area corresponds to their Perceptual system. In addition, our architecture includes a further level (the conceptual area), which is
Conclusions
In the above sections, a possible interpretation of the language of the situation calculus in terms of conceptual spaces is suggested. In this way, the situation calculus can be adopted as the formalism for the linguistic area of the model, with the advantage of using a powerful, well understood and widespread formal tool.
In recent years, in the field of cognitive robotics, various formalisms based on situation calculus have been proposed, such as [12], [13], [21]. They allow the programmer to
Acknowledgements
Thanks to Gigina Aiello, Edoardo Ardizzone, Ron Arkin, Christian Balkenius, Shimon Edelman, Peter Gärdenfors, Donatella Guarino, Lars Kopp, Yves Lesperance, Fiora Pirri, Ray Reiter, Alessandro Saffiotti, Pino Spinelli and Luc Steels for having discussed with us over the years, the topics of this paper. Ignazio Infantino and Francesco Raimondi contributed to the hardware and software implementation of the system. This work has been partially supported by the Project “Efficient robot vision
A. Chella was born in Florence, Italy, on 4 March 1961. He received his laurea degree in Electronic Engineering in 1988 and his Ph.D. in Computer Engineering in 1993 from the University of Palermo, Italy. Since 2001 he is a professor of robotics at the University of Palermo and a scientific advisor of the CERE (Center of Study of Computer Networks) of the Italian Research, Council (CNR). His research interests are in the field of autonomous robotics, artificial vision, hybrid
References (33)
- et al.
Experiences with an interactive museum tour-guide robot
Artificial Intelligence
(1999) - et al.
A cognitive architecture for artificial vision
Artificial Intelligence
(1997) - et al.
An architecture for autonomous agents exploiting conceptual representations
Robotics and Autonomous Systems
(1998) - et al.
Understanding dynamic scenes
Artificial Intelligence
(2000) - et al.
ConGolog, a concurrent programming language based on the situation calculus
Artificial Intelligence
(2000) The spatial semantic hierarchy
Artificial Intelligence
(2000)Prolegomena to a theory of mechanized formal reasoning
Artificial Intelligence
(1980)- R. Arkin, Behavior-based Robotics, MIT Press, Cambridge, MA,...
- A. Blake, M. Isard, Active Contours, Springer, Berlin,...
- A. Chella, M. Frixione, S. Gaglio, Conceptual spaces for symbol anchoring—review of the book: Conceptual Spaces. The...
Teleautonomous systems: projecting and coordinating intelligent actions at a distance
IEEE Transactions on Robotics and Automation
Cited by (0)
A. Chella was born in Florence, Italy, on 4 March 1961. He received his laurea degree in Electronic Engineering in 1988 and his Ph.D. in Computer Engineering in 1993 from the University of Palermo, Italy. Since 2001 he is a professor of robotics at the University of Palermo and a scientific advisor of the CERE (Center of Study of Computer Networks) of the Italian Research, Council (CNR). His research interests are in the field of autonomous robotics, artificial vision, hybrid (symbolic–subsymbolic) systems and knowledge representation. He is a member of IEEE, ACM and AAAI.
M. Frixione was born in Genoa, Italy, in 1960. He received his Laurea degree and his Ph.D. in Philosophy from the University of Genoa, respectively, in 1986 and 1993. Currently, he is Assistant Professor in Philosophy of Language at the Department of Communication Sciences of the University of Salerno, Italy. His research interests are in the field of Cognitive Sciences and Artificial Intelligence, and include Knowledge Representation, Hybrid Systems and the philosophical aspects of Cognitive Sciences.
S. Gaglio was born in Agrigento, Italy, on 11 April 1954. He graduated in electronic engineering at the University of Genoa, Genoa, Italy in 1977. In 1977 he was awarded a Fulbright scholarship to attend graduate courses in USA, and in 1978 he received the M.S.E.E. degree from the Georgia Institute of Technology, Atlanta, USA. Since 1986 he is a professor of artificial intelligence at the University of Palermo, Italy. Since 1999 he is the Director of CERE (Center of Study of Computer Networks) of the Italian Research Council (CNR). He has been member of various committees for projects of national interest in Italy and he is referee of various international scientific congresses and journals. His present research activities are in the area of artificial intelligence and robotics. He is a member of IEEE, ACM, and AAAI.