Activities are located behaviors, taking time, conceived as socially meaningful, and usually involving interaction with tools and the environment. In modeling human cognition as a form of problem solving (goal-directed search and operator sequencing), cognitive science researchers have not adequately studied "off-task" activities (e.g., waiting), non-intellectual motives (e.g., hunger), sustaining a goal state (e.g., playful interaction), and coupled perceptual-motor dynamics (e.g., following someone). These aspects of human behavior have been considered in bits and pieces in past research, identified as scripts, human factors, behavior settings, ensemble, flow experience, and situated action. More broadly, activity theory provides a comprehensive framework relating motives, goals, and operations. This paper ties these ideas together, using examples from work life in a Canadian High Arctic research station. The emphasis is on simulating human behavior as it naturally occurs, such that "working" is understood as an aspect of living. The result is a synthesis of previously unrelated analytic perspectives and a broader appreciation of the nature of human cognition. Simulating activities in this comprehensive way is useful for understanding work practice, promoting learning, and designing better tools, including human-robot systems.

In this paper we present an abstract agent programming language and its operational semantics which can be used to implement cognitive agents. This language consists of programming constructs to implement both the agent's mental attitudes -- interpreted as data structures -- as well as the agent's deliberation process. The agent can observe the environment, generate goal sets from desires, obligations, and intentions, selects goals, generate plans, and execute them. These actions can be combined in the deliberation language in a variety of ways to program the agent's deliberation process. At the level of abstraction of our deliberation language, goal generation and planning are both characterized as conflict resolution procedures. For goal generation, obligation, desire and intention rules can conflict when the corresponding goals are incompatible. For planning, partial plans can be incompatible. In our approach, the incompatibility of plans can be derived from more detailed data structures such as resources of the agents, but the conflict procedure can also be programmed directly by the agent programmer.

Abstract. This paper explores the design and implementation of a societal arrangement of reflexive and motivational agents which will act as the building blocks for a more abstract agent within which the current agents act as distributed dynamic processing nodes. We contest that reactive, deliberative and other behaviours are required in complete (intelligent) agents. We provide some architectural considerations on how these differing forms of behaviours can be cleanly integrated and relate that to a discussion on the nature of motivational states and the mechanisms used for making decisions. 1

In this article, we present an isotropic unsupervised algorithm for temporal sequence learning. Nospecial reward signal is used such that all inputs are completely isotropic. All input signals are bandpass filtered before converging onto a linear output neuron. All synaptic weights change according to the correlation of bandpass-filtered inputs with the derivative of the output. We investigate the algorithm in an open- and a closed-loop condition, the latter being defined by embedding the learning system into a behavioral feedback loop. In the open-loop condition, we find that the linear structure of the algorithm allows analytically calculating the shape of the weight change, which is strictly heterosynaptic and follows the shape of the weight change curves found in spike-time-dependent plasticity. Furthermore, we show that synaptic weights stabilize automatically when no more temporal differences exist between the inputs without additional normalizing measures. In the second part of this study, the algorithm is is placed in an environment that leads to closed sensormotor loop. To this end, a robot is programmed with a prewired retraction reflex reaction in response to collisions. Through isotropic sequence order (ISO) learning, the robot achieves collision avoidance by learning the correlation between his early range-finder signals and the later occurring collision signal. Synaptic weights stabilize at the end of learning as theoretically predicted. Finally, we discuss the relation of ISO learning with other drive reinforcement models and with the commonly used temporal difference learning algorithm. This study is followed up by a mathematical analysis of the closed-loop situation in the companion article in this issue, “ISO Learning Approximates a Solution to the Inverse-Controller Problem in an Unsupervised Behavioral Paradigm” (pp. 865–884).

Two layer control systems are common in robot architectures. The lower level is designed to provide fast, fine grained control while the higher level plans longer term sequences of actions to achieve some goal. Our approach uses reinforcement learning (RL) for the low level and Partially Observable Markov Decision Process (POMDP) planning for the high level. Because both levels can adapt their behavior within the scope of their tasks, the combination is expected to be robust to degradations in sensor and actuator failures and so to enhance overall system reliability. We implemented our architecture for use in the Khepera robot simulator. In a set of experiments, we show that good performance can be difficult to achieve with hand coded low level control and that performance of our RL/POMDP system degrades slowly with increasing sensor and actuator failure. 1 Introduction Two layer control systems are common in robot architectures. Two layer control is a pragmatic solution to different...

An autonomous agent operating in a dynamical environment will face a number of different reasoning problems, one of which is how to plan its actions in order to pursue its goals. For this purpose, it is important that the agent represents its knowledge about the world in a coherent, expressive and well-understood way, in our case the temporal logics from Erik Sandewall's "Features and Fluents" framework. However, most existing planning systems make no use of temporal logics, but have specialised representations such as the Strips formalism and hierarchical task networks. In order to benefit from the techniques used by these planners, it is useful to analyse and reconstruct them within the given framework. This includes making explicit the ontological and epistemological assumptions underlying the planners; representing plans as entities of the temporal logic; and reconstructing the algorithms in terms of the new representation. The two planners Strips and Tweak have been analysed and r...

this paper on these three: representation, action, and motivation. In particular, I will argue that the standard view of representation as some kind of correspondence, as an encoding, is wrong. I outline an alternative model of representation that emerges naturally in agents, biological or designed, that actually engage the world (Beer, 1990, 1995, in press; Beer, Chiel, Stirling, 1990; Bickhard, 1980, 1993; Bickhard & Terveen, 1995; Brooks, 1991a, 1991b, 1991c; Cherian & Troxell, in press; Malcolm, Smithers, Hallam, 1989; Smithers, 1994). One primary consequence of this alternative model of representation --- called interactivism --- is that functions that are standardly taken to reside in separate modules, such as representation, action, and motivation, are inherently integrated as separate functional aspects of one single underlying ontology. They are not inherently distinct modules. If standard models that permit such modularization are in error, then so are such modularizations per se. 2 Encoding Models of Representation.

The paper presents a general architecture for behaviour based control systems for autonomous agents. A number of archi tectural principles are proposed which make it possible to combine reactive control with learning and problem solving in a coherent way. In particular, I investigate the interaction between reinforcement learning, internal world models and dynamic action selection as well as a number of connections to psychological models and biological systems.

Abstract. A new approach to building robots with anticipatory behavior is presented. This approach is based on analogy with a single episode from the past experience of the robot. The AMBR model of analogy-making is used as a basis, but it is extended with new agent-types and new mechanisms that allow anticipation related to analogical transfer. The role of selective attention on retrieval of memory episodes is tested in a series of simulations and demonstrates the context sensitivity of the AMBR model. The results of the simulations clearly demonstrated that endowing robots with analogy-based anticipatory behavior is promising and deserves further investigation. 1

Abstract. Autonomic networking set a challenge for the research community to engineer systems and architectures that will increase the QoS and robustness of future network architectures. However, our experience is that so far the autonomic network research community does not have a common perception of what an autonomic network is. This paper attempts to propose a generic model for autonomic systems, along with a minimum set of required properties that would render a system compliant to this model. The paper emphasises the importance of such a common model for the credibility of the research community as well as to eliminate attempts to unnecessarily overload or blur the scope of the field.