Modeling robot cognitive activity through active mental entities
Introduction
Cognitive issues have been receiving increasing attention in the field of advanced robotics [13], [49], [61]. High-level desiderata concerning the capabilities of an autonomous robot are often expressed in terms of general cognitive properties: for instance, the issues of intelligence and autonomy are discussed in [61], whereas Brooks [13] suggests a shift from behavior-based to cognitive robotics and insists on robot motivation and coherence.
While there is a general understanding that cognitive issues may play a profitable role both in the specification and in the evaluation of robot performances, there is no general agreement about how to equip a robot with suitable forms of cognition. Some researchers suggest that an explicit representation of cognitive activity should be encompassed within robot architectural design, whereas others conceive cognition as an emerging property, which can be achieved by exploiting the dynamics of system–environment interaction, without the need of explicit internal cognitive structures [49], [50].
In the former case there is the problem of defining the structure and operation of the cognitive level and of integrating it with other components of the control architecture, whereas in the latter case the robot designer has to devise internal operation mechanisms able to produce the desired cognitive features, at the level of external behavior. Both approaches clearly feature their own advantages and drawbacks.
In this paper we propose an original approach to the design and realization of cognitive issues which aims at overcoming the opposition between the two traditional standpoints that advocate “explicit representation” and “emerging behavior”. Starting from an analysis of the design dimensions of an autonomous robot architecture, we show that, differently from the other ones, the motivational dimension has not yet fully benefited from the evolution towards a distributed organization that can be noticed in recent literature. We therefore propose a model of robot motivations based on the original notion of active mental entity: robot mental attitudes are explicitly represented as autonomous computational entities so that the overall control architecture can be conceived as a community of interacting entities, including mental attitudes and sensory and actuation devices. Such explicit model is coherent with the distributed organization of other architecture dimensions. It is however also in accordance with the idea of “emerging intelligence”, since mental activity can be understood as the emerging result of the interactions involving both active mental entities and the external environment. Though being related to several recent works in the area of robot control architectures, the proposed approach shows several unique features and is characterized by clear relationships between its internal organization and the cognitive issues it is intended to realize.
The paper is organized as follows. In Section 2 we survey and discuss the evolution of robot control architectures in recent years. In Section 3 we provide some arguments supporting the importance of endowing an autonomous robot with a motivational structure. Section 4 introduces and discusses the notion of active mental entity, while in Section 5 two main classes of active mental entities, namely intenders and attenders, are described. In Section 6 the application of the overall architectural scheme in a simulated experimental context is presented. Section 7 analyzes the relations between our proposal and other related approaches, whereas Section 8 summarizes and concludes the paper.
Section snippets
The evolution of autonomous robot control architectures: A survey
From a historical perspective, the evolution of autonomous robot architectures is well known: early approaches, based on a strict hierarchical paradigm, functional decomposition, and the so-called Good Old Fashioned Artificial Intelligence [13] (GOFAI) techniques, proved to be fairly inadequate in practice. Their failure paved the way to behavior-based approaches [10], characterized by subsumption architecture, task-based decomposition, and reactive stimulus-response mechanisms.
Behavior-based
Why modeling motivations for autonomous robot control?
As mentioned in the previous section, the issue of embedding motivations within an autonomous robot is a further important dimension in the architecture design.
As shown in Table 1, two main approaches can be found in literature. On one side, behavior-based school of thought excludes any representation of motivations, as well as of mental attitudes of any kind, since “intelligence emerges from the interaction of the components of the system” [11]. On the other side, most approaches allow the
Active mental entities: A new approach to modeling robot mental activity
Mental activity is often described by using terms such as desire, belief, intention, hope, obligation, prohibition, etc. These terms denote mental entities, i.e. entities that are inside the mind of an intelligent agent and which are responsible of his/her external behavior. Mental entities therefore represent the basic motivations driving an autonomous agent.
Our main point concerns the fact that mental activity should be modeled by providing mental entities with a sort of “agentification”. In
Intenders and attenders
In the following, the main features of intenders and attenders are outlined. This section does not aim at describing the technical details concerning the definition of such active mental entities, it rather focuses on those characteristics and operation mechanisms which prove to be crucial for realizing two significant cognitive issues related to motivation, namely coherence and autonomy.
By coherence we mean the capability of establishing some general and stable points of reference guiding
Embedding active mental entities within the Khepera simulator
In this section, we illustrate how the proposed approach to modeling agent mental activity can be suitably exploited for the realization of the control system of an autonomous mobile robot.
In particular, a software prototype implementing the proposed paradigm has been developed and used to realize a multi-agent control architecture for the Khepera robot simulator [39]. Some experimental results are provided in order to illustrate the operation of such articulated control structure in a
Discussion and comparison
In this section we will discuss the relationships between our proposal and the existing literature, by examining three main aspects: the overall architecture organization, the representation of motivation, and the operation scheme of active mental entities.
Conclusions
While cognitive issues have been receiving an increasing attention in the autonomous robotics field, the issue of how to equip robots with motivation and cognition is still rather debated.
In this paper, building on the standpoint that logical and computational distribution are sound design principles to be applied also to motivational models, we have proposed an original approach to model robot mental activity based on the concept of active mental entities, i.e. on the representation of mental
Acknowledgements
The authors are indebted to the anonymous referees for their insightful comments and to Professor Giovanni Guida for his support and for many useful discussions about the topics of this paper.
The authors gratefully thank Dr. Stefania Ruffini for her help in improving the quality of the text.
Pietro Baroni was born in Brescia, Italy, in 1966. He received the “Maturità Classica” degree from Liceo Bellini-Pastore, Castiglione delle Stiviere, Italy, the “Laurea” degree in Mechanical Engineering, from University of Brescia, Italy, and a “Master” degree in Information Technology from CEFRIEL, Milan, Italy. Since 1993 he is an Assistant Professor at the Faculty of Engineering, University of Brescia, Department of Electronics for Automation, where he works in the Knowledge Engineering and
References (66)
Intelligence without representation
Artificial Intelligence
(1991)From earwigs to humans
Robotics and Autonomous Systems
(1997)- et al.
An architecture for autonomous agents exploiting conceptual representation
Robotics and Autonomous Systems
(1998) An assumption-based TMS
Artificial Intelligence
(1986)A truth maintenance system
Artificial Intelligence
(1979)An architecture for adaptive intelligent systems
Artificial Intelligence
(1995)- et al.
A parallel processing architecture for sensor-based control of intelligent mobile robots
Robotics and Autonomous Systems
(1996) Controlling cooperative problem solving in industrial multi-agent systems using joint intentions
Artificial Intelligence
(1995)- et al.
Adaptive selection of reactive/deliberate planning for a dynamic environment
Robotics and Autonomous Systems
(1998) - et al.
Planning robust displacement missions by means of robot-tasks and local maps
Robotics and Autonomous Systems
(1997)
Cognition — Perspectives from autonomous agents
Robotics and Autonomous Systems
Sensory-motor cordination: The metaphor and beyond
Robotics and Autonomous Systems
The uses of plans
Artificial Intelligence
Justification and defeat
Artificial Intelligence
When are robots intelligent autonomous agents?
Robotics and Autonomous Systems
A behavior-based blackboard architecture for reactive and efficient task execution of an autonomous robot
Robotics and Autonomous Systems
An architecture for autonomy
International Journal of Robotics Research
Callisto — A multi-agent robot trash-collecting team
AI Magazine
A retrospective of the AAAI robot competitions
AI Magazine
Experiences with an architecture for intelligent reactive agents
Journal of Experimental and Theoretical Artificial Intelligence
Cited by (5)
Development of intelligent multisensor surveillance systems with agents
2007, Robotics and Autonomous SystemsCitation Excerpt :This coordination will improve system functioning, for instance, in the evaluation of the relevance of events that are captured by several sensors; by being able to track elements moving around the system under surveillance; or for the collaboration of several effectors to solve some problem. A way of implementing decentralization, autonomy and coordination needs is by means of agent technology [2,4]. From the point of view of this technology, intelligent multisensor surveillance systems would be considered as multiagent systems (MAS).
An Overview on the Development and Application of Cognitive Robot
2022, Frontiers in Artificial Intelligence and ApplicationsA level task allocation method for multiple-agents
2013, Journal of Computational Information SystemsArgumentation through a distributed self-stabilizing approach
2002, Journal of Experimental and Theoretical Artificial IntelligenceA distributed self-stabilizing algorithm for argumentation
2001, Proceedings - 15th International Parallel and Distributed Processing Symposium, IPDPS 2001
Pietro Baroni was born in Brescia, Italy, in 1966. He received the “Maturità Classica” degree from Liceo Bellini-Pastore, Castiglione delle Stiviere, Italy, the “Laurea” degree in Mechanical Engineering, from University of Brescia, Italy, and a “Master” degree in Information Technology from CEFRIEL, Milan, Italy. Since 1993 he is an Assistant Professor at the Faculty of Engineering, University of Brescia, Department of Electronics for Automation, where he works in the Knowledge Engineering and Human-Computer Interaction research group. His research interests include agent architectures, multi-agent systems, uncertain reasoning, and automated diagnosis.
Daniela Fogli was born in Pesaro, Italy, in 1970. She received the Laurea degree in Computer Science from the University of Bologna, Italy, in 1994 and the Ph.D. degree in Information Engineering from the University of Brescia, Italy, in 1998. She is currently a post-doc grant holder at the Joint Research Centre of the European Commission, Institute for Systems, Informatics and Safety. Her research interests are concerned with agents and multi-agent systems, autonomous mobile robot control architectures, knowledge-based systems, and software dependability assessment.