Showing papers on "Cognitive robotics published in 1990"

Complex systems and cognitive processes

Abstract: This book shows that the science of complex systems, which stresses the importance of self-organizing processes, can make a decisive contribution to solving many problems in artificial intelligence. Artificial cognitive systems are important in view of their potential applications, and it can be expected that their study will shed light on biological cognitive systems. The new "neurally inspired" information science proposed in this book is fast becoming a promising workshop for the construction of models capable of emulating cognitive behaviour. After a general introduction to the theory of complex systems, the book gives a thorough treatment of neural networks, which are the most successful and the most thoroughly studied dynamical cognitive systems. Attention is also devoted to other classes of artificial cognitive systems, in particular to classifier systems, which provide an important link between the dynamical and the inferential approach to artificial intelligence. The book can be used as a textbook, since it does not require previous knowledge of the topic, and should also be interesting for researchers in this field, since it links formerly separate lines of research.

Memory-Based Learning in Intelligent Control Systems

Abstract: Memory-based learning is becoming more popular in Artificial Intelligence and can be used in Intelligent Control Systems. This talk will describe how an associative content-addressable memory can be used to model a robot and the world the robot interacts with. The model can be learned by storing experiences in the memory. To make predictions the memory is searched for relevant experience. An initial implementation of such a memory-based modeling scheme has been made on a parallel computer, the Connection Machine. The implementation was used to model and control a simulated planar two-joint arm and a simulated running machine. This paper describes the issues and problems that arose in this preliminary work (Atkeson and Reinkensmeyer 1989).

A computational model for a robotic arm instructed by natural language

Abstract: An effort to develop a robotic system that can be instructed by a nonexpert user in an office environment using natural language is described. The central problem is the transformation of ordinary English utterances into unambiguous task structures that can be performed by a robot. The research reveals that a robotic system using natural language must include robot-level programming strategies if it is to be efficient. Moreover, to facilitate and enhance communication between a user and a robot, the robot-level programming language needs to be designed to reflect the way humans refer to actions, and to reflect temporal and logical relationships expressed in natural language. In place of a traditional point-to-point robot control scheme, a concurrent, perception-feedback-driven control strategy has been developed. An initial book handling experiment has been conducted in an officelike environment. >

Control and valuation in human cognition

Abstract: A theoretical model of human cognition that attempts to explain how cognition is controlled and how humans determine what is important is outlined. There are three concepts critical to the model: motivational states (goals) structure and control all cognitive processes; a top-level mechanism controls and integrates processing of conscious and unconscious streams of cognition; and goals are selected for processing after valuation by up to four methods: reward system, drives, emotions, and rationality. The model can account for many phenomena from the cognitive psychology literature including motivational states, activation, attention, and consciousness. The model is compared with competing theories. >