View metadata, citation and similar papers at core.ac.uk brought to you by CORE QUARTERLY REVIEW No.20 / July 2006 1 Cognitive Robotics to Understand Human Beings KAYOKO ISHII Life Science Unit taking place at that time. 1 Introduction In Japan, research on the computational theory of the brain[1] and research combining The question of what human beings (self theory and physiological experiments[2] has been and others), the mind, and the world are has carried out. One of the major themes in Japanese always been of great interest to humankind. brain science in the 1990s was “creating the Most academic disciplines originated to answer brain” beside analytical experimental sciences these questions. As neuroscience has advanced, (“understanding the brain”) and research oriented the notion that brain function is closely related to medical applications (“protecting the brain”). to the mind has become more widely accepted, This theme was significant in that it not only increasing the expectation that unknown aspects expressed the concept of understanding brain of the mind could be explored by neuroscience. functions through “cycles of creation of models of However, a long-standing question regarding the brain, computational theory and neural networks, mind is that one’s mind seems to be associated their verification through experimental science, with oneself as a physical existence, yet its and improvement of theories and models,” but content seems not to be expressed physically. also expressed the unconventional orientation Simply accumulating knowledge acquired of creating new systems inspired by the brain. through external analysis and observation of Furthermore, computational neuroscience was the brain as a physical entity is not sufficient to defined as “to investigate information processing elucidate the essential functions of the brain and of the brain to the extent that artificial machines, the nature of the mind. either computer programs or robots, can solve The brain has many parallel units (modules) the same computational problems as solved by that represent different parts of the body the brain, essentially in the same principle” [3]. or participate in different functions. When Based on this conceptual framework, innovative neuroscientists study the properties of a module, researchers, although still few in number, are they apply a controlled stimulus to the subject, engaging in studies to elucidate human brain so that it perturbs only the targeted module functions through “cycles of creation of brain (or limited numbers of modules, including the algorithms, their verification through robots, target). When human beings engage in usual noninvasive measurements of brain activities, activities, however, many different modules psychology, and experimental sciences, and work in an autonomous and distributed manner. further improvement of the algorithms.” Particular ideas or actions are generated either From the perspective of ordinary Japanese by the exchange of information between sensibilities as well, the mind cannot be specific modules or the selective involvements considered in isolation from the body, the of certain modules. Unless the algorithms for environment, and the existence of other people. these information exchanges and selections can In other words, attempting to create the brain be elucidated, observation of the physical state alone will not elucidate the essential functions of of the brain at a given time cannot lead to an the brain itself or the mechanisms of the mind. understanding of the information processing Embodiment and context dependence are key 11 SCIENCE & TECHNOLOGY TRENDS QUARTERLY REVIEW No.20 / July 2006 concepts in cognitive science and neuroscience, 2-1 Cognitive robotics and robots given bodies to interact with the “Cognitive robotics” in this report refers to environment are serving as effective simulation an comprehensive science in which robotics, tools [4]. as described above, neurosciences (ranging from the experimental to the theoretical or mathematical variety and neuroinformatics), 2 The Field of Cognitive cognitive science, psychology (psychophysics and Robotics behavioral measurement) and behavioral sciences Since their beginning, robots have been seamlessly collaborate in unity while keeping constructed to imitate, replace, and supplement variations in perspective, closely connecting human beings or a part of human functions. with fields such as philosophy, social sciences, Since 1960, the mainstream of robot development anthropology, and economics; exchanging has been oriented to industrial applications - their knowledge and methodologies, and manufacturing robots. In recent years however, executing mutual verification. Robotics herein we have witnessed a rapid increase in the represents an expectation of interdisciplinary development of robots designed to serve ordinary integration, rather than the simple collection people rather than experts [5 -7]. Traditionally, of independent research fields, and will be robotics referred to a combination of science, realised by using robots as a common verification engineering, psychology, sociology, and other platform to highlight weaknesses and errors in disciplines necessary “for the development, research processes in individual disciplines and construction, and dissemination of practical contradictions among different disciplines, robots,” with particular emphasis on the For the development of commercial robots, a engineering aspects. demand-oriented perspective, based on future During the process of seeking the necessary prospects and on a broad understanding of conditions for robots to act as flexibly, smoothly, humans and society, is required. For humanoid and autonomously as human beings in the real robots (humanoids hereafter), the initial phase of world, robotics researchers began to turn their development has been completed for structural attention to human cognitive mechanisms, modules and actuators, which correspond to learning, recognition of others, and social their bodies, and the computers to control them. behaviors. In Japan, since around 1994, robotics Currently, humanoid hardware developed in researchers have organized research groups, Japan is widely used, both in Japan and abroad such as the Keihanna Research Group for as platforms for the development of software to Sociointelligenesis, with the primary aim of serve as the cognitive mechanisms. Certainly, elucidating human cognition, development, and future success in the development of practical behaviors by using robots. These researchers humanoids will depend on the improvement of have adopted a “constructivist” approach, which their “cognitive functions.” aims to explain human cognitive mechanisms Some research laboratories dedicated to the by creating and testing robots that can develop development of practical robots have taken up humanlike cognitive abilities to cope with the parallel research in cognitive robotics, which real world (“cognitive developmental robotics”)[8]. offers this fundamental knowledge. Even These researchers have the advantage of having researchers specializing in the development not only advanced knowledge and experiences in of structural modules and actuators must take physics and engineering, which share a common compatibility with next-generation cognitive basis in mathematics, but also a broad knowledge functions into consideration. Energy consumption and understanding of biology, the humanities, is another important future issue expected and the social sciences, and of having a solid to accompany advances in robotic cognitive verification platform, such as robots. function. Research to solve this issue will become necessary. 12 13 SCIENCE & TECHNOLOGY TRENDS QUARTERLY REVIEW No.20 / July 2006 Figure 1 : Cognitive robotics Mathematics Philosophy Computational neur Social Theor osciences Theoreticalysics sciences ph y or Economics iented Neur inf Anthro- ormaticso- pology ing Mater vioral mation technology ial or or Beha sciences Inf iented Psychology System engineer Mechanics Cognitive Neuro- System biology sciences sciences Materials Biology Prepared by the STFC Energy consumption problems associated with improvement in cognitive functions Neural systems are enormous energy consumers in animals. Humans have remarkably large brains for their bodies (approximately 2.5% of body weight), and the brain consumes 20% of the body’s total energy. The cerebral neocortex of primates increased exponentially in volume as their social behaviour became more complex (the social brain hypothesis, Reference 14). In order for robots to work in complex human society, improvement of their cognitive functions is indispensable. This raises the problem of energy consumption for information processing. Furthermore, because of size restrictions and mobility requirements for humanoid robots, it is critical to invent new materials and structures enabling flexible and efficient information processing within limited spaces. This means that humanoid robots can be most desirable test beds for the creation of new paradigms for computational theories and materials/structures. Although noninvasive methods to measure Such algorithms, if available, would serve as brain activity have progressed, certain constraints the foundation for the development of new remain. Currently, interpretation of results computers and human-machine interfaces. requires either (1) statistical analysis of data from multiple measurements or (2) training subjects 2-2 Global progress of robotics