Abstract. To explore issues of developmental structure, physical embodiment, integration of multiple sensory and motor systems, and social interaction, we have constructed an upper-torso humanoid robot called Cog. The robot has twenty-one degrees of freedom and a variety of sensory systems, including visual, auditory, vestibular, kinesthetic, and tactile senses. This chapter gives a background on the methodology that we have used in our investigations, highlights the research issues that have been raised during this project, and provides a summary of both the current state of the project and our long-term goals. We report on a variety of implemented visual-motor routines (smooth-pursuit tracking, saccades, binocular vergence, and vestibular-ocular and opto-kinetic reflexes), orientation behaviors, motor control techniques, and social behaviors (pointing to a visual target, recognizing joint attention through face and eye finding, imitation of head nods, and regulating interaction through expressive feedback). We further outline a number of areas for future research that will be necessary to build a complete embodied system. 1

We present a novel methodology for building humanlike artificially intelligent systems. We take as a model the only existing systems which are universally accepted as intelligent: humans. We emphasize building intelligent systems which are not masters of a single domain, but, like humans, are adept at performing a variety of complex tasks in the real world. Using evidence from cognitive science and neuroscience, we suggest four alternative essences of intelligence to those held by classical AI. These are the parallel themes of development, social interaction, embodiment, and integration. Following a methodology based on these themes, we have built a physical humanoid robot. In this paper we present our methodology and the insights it affords for facilitating learning, simplifying the computation underlying rich behavior, and building systems that can scale to more complex tasks in more challenging environments.

The evaluation of models of social and behavioral development is difficult in natural settings; ethical concerns, difficulties in implementing experimental procedures, and difficulties in isolating hypothesized variables make experimental evidence difficult or impossible to obtain. We propose the use of human-like robots as a testbed for the evaluation of models of human social development. Robotic implementation of human social models allows for unique opportunities to evaluate those models. In this paper, we review some of the implications of this proposal by examining a case study of an on-going project to implement an existing model of one aspect human social development, the development of joint attention behaviors.

This chapter highlights the importance of the problem of action coding, that is, the cognitive representation of action, for theories of S-R compatibility. An action-concept model of S-R compatibility is presented, based on considerations of Lotze and HarleB on the emergence of voluntary action. It assumes that the cognitive code of any perceivable, movement-contingent event--hence, action effectmis associated with the motor pattern producing it. Accordingly, the cognitive system can, and actually does, use these action-effect codes to choose between actions and to address motor patterns for action generation. That is, acton-effect codes serve for perception as well as for action control, and are thus called action concepts. The explanatory power of the action-concept model is demonstrated for a considerable number of findings from compatibility research. It is argued that such a model could close a theoretical gap in understanding S-R compatibility and the perception-action relationship in general. Introduction: The Problem of Action Coding

This paper presents an approach to robot arm control based on exploiting the dynamical properties of an adaptive oscillator circuit coupled to the joints of an arm. The approach is implemented on a real robot arm, and swings pendulums at their natural frequencies, turns cranks and manipulates slinky toys. These actions are all achieved using the same architecture, without any modeling of the arm or its environment. The simple nature of the oscillators, and the lack of modeling results in a robust and very simple system. 1 Introduction This paper presents an approach to the control of robot arms which exploits the physical coupling of the arm and its environment. Dynamical characteristics of the arm are exploited rather than modeled, allowing a very simple control scheme to exhibit a variety of interesting rhythmic behaviors. The system is implemented on the arms of the humanoid robot Cog [ Brooks and Stein, 1994 ] , which is illustrated in Figure 1. Robot arms are complex systems, wit...

We propose that connectionism and dynamic systems theory are strong contenders for a general theory of development that holds true whatever the content domain. We illustrate, through our own career narratives, the origins of these theories in motor and language development. We situate connectionism and dynamic systems among other classic and contemporary theories and conclude that, although there are meaningful differences, these differences pale in relation to the shared assumptions about the fundamental processes and mechanisms of change.

Humans exploit dynamics---gravity, inertia, joint coupling, elasticity, and so on---as a regular part of skillful, coordinated movements. Such movements comprise everyday activities, like reaching and walking, as well as highly practiced maneuvers as used in athletics and the performing arts. Robots, especially industrial manipulators, instead use control schemes that ordinarily cancel the complex, nonlinear dynamics that humans use to their advantage. Alternative schemes from the machine learning and intelligent control communities offer a number of potential benefits, such as improved efficiency, online skill acquisition, and tracking of nonstationary environments. However, the success of such methods depends a great deal on structure in the form of simplifying assumptions, prior knowledge, solution constraints and other heuristics that bias learning. My premise

This paper presents an approach to robot arm control based on exploiting the dynamical properties of a simple oscillator circuit coupled to the joints of an arm. Using the same architecture, a wide variety of tasks can be achieved without any explicit modeling of the arm or its environment. The inherent properties of the oscillators give robustness to perturbations and changes in frequency. The approach is implemented on two compliant arms, and is demonstrated to swing pendulums at their natural frequencies, turn cranks and manipulate 'Slinky' toys.

Trajectory information can be analysed in both the time and space dimensions via a new metric called the 7/t-metric. The 7/t-metric is a measurement definition which quantifies the allocation of control across multiple degrees of freedom. Allocation of control is defined as the product of two components, the simultaneity of control and the efficiency of control, corresponding to the time and space dimensions respectively. The existing human factors, biomedical, and motor control literature serves as the foundation for the development of the 7/t- metric. The 7/t-metric has several limitations including dependency upon the chosen coordinate system, assumptions of optimal trajectories, and the lack of frequency domain analysis.

This paper presents two complementary ideas relating the study of human development and the construction of intelligent artifacts. First, the use of developmental models will be a critical requirement in the construction of robotic systems that can acquire a large repertoire of motor, perceptual, and cognitive capabilities. Second, robotic systems can be used as a test-bed for evaluating models of human development much in the same way that simulation studies are currently used to evaluate cognitive models. To further explore these ideas, two examples from the author's own work will be presented: the use of developmental models of hand-eye coordination to simplify the task of learning to reach for a visual target and the use of a humanoid robot to evaluate models of normal and abnormal social skill development. Introduction Research on human development and research on the construction of intelligent artifacts can and should be complementary. Studies of human developm...