CHAPTER 1
INTRODUCTION
This document is the final report of a study on the feasibility of using machine intelligence, including automation and robotics, in future space missions. The 10-week study was conducted during the summer of 1980 by 18 educators from universities throughout the United States who worked with 15 NASA program engineers. The specific study objectives were to identify and analyze several representative missions that would require extensive applications of machine intelligence, and then to identify technologies that must be developed to accomplish these types of missions.
The study was sponsored jointly by NASA, through the Office of Aeronautics and Space Technology (OAST) and the Office of University Affairs, and by the American Society for Engineering Education (ASEE) as part of their continuing program of summer study faculty fellowships. Co-hosts for the study were the NASA-Ames Research Center and the University of Santa Clara, where the study was carried out. Project co-directors were James E. Long of the Jet Propulsion Laboratory and Timothy J. Healy of the University of Santa Clara.
The study was sponsored by NASA because of an increasing realization of the major role that advanced automatic and robotic devices, using machine intelligence, must play in future space missions (fig. 1.1). Such systems will complement human activity in space by accomplishing tasks that people cannot do or that are otherwise too dangerous, too laborious, or too expensive. The opportunity to develop a powerful new merger of human intellect and machine intelligence is a result of the growing capacity of machines to accomplish significant tasks. The study has investigated some of the ways this capacity may be used as well as a number of research and development efforts necessary in the years ahead if the promise of advanced automation is to be fully realized.
1.1 Survey of Artificial Intelligence
Many of the concepts and technologies considered in this study for possible use in future space missions are elements of a diverse field of research known as "artificial intelligence" or simply AI. The term has no universally accepted definition or list of component subdisciplines, but is commonly understood to refer to the study of thinking and perceiving as general information processing functions - the science of intelligence. Although, in the words of one researcher, "It is completely unimportant to the theory of AI who is doing the thinking, man or computer" (Nilsson, 1974), the historical development of the field has followed largely an empirical and engineering approach. In the past few decades, computer systems have been programmed to prove theorems, diagnose diseases, assemble mechanical equipment using a robot hand, play games such as chess and backgammon, solve differential equations, analyze the structure of complex organic molecules from mass-spectrogram data, pilot vehicles across terrain of limited complexity, analyze electronic circuits, understand simple human speech and natural language text, and even write computer programs according to formal specifications - all of which are analogous to human mental activities usually said to require some measure of "intelligence." If a general theory of intelligence eventually emerges from the AI field, it could help guide the design of intelligent machines as well as illuminate various aspects of rational behavior as it occurs in humans and other animals.
AI researchers are the first to admit that the development of a general theory of intelligence remains more a goal for the future than an accomplishment of the present. In the meantime, work is progressing in a number of more limited subdisciplines. The following seven topical research areas include most elements normally considered to be a part of the field.
1.1.1 Planning and Problem Solving
All of artificial intelligence involves aspects of planning and problem solving, a rather generic category. This includes planning and organization in the program development phase as well as the dynamic planning required during an actual mission. Problem solving implies a wide range of tasks including decision making, optimization, dynamic resource allocation, and many other calculations or logical operations that arise throughout a mission.
1.1.2 Perception
Perception is the process of obtaining data from one or more sensors, and analyzing or processing these data to facilitate subsequent decisions or actions. One simple example is a visual perception system that views a scene, determines whether or not a specified round object is in the scene, and if so, initiates a signal which causes an automatic arm to move the object out of the scene. Perception may be
