"In this paper we describe some major new additions to the STRIPS robot problem-solving system. The first addition is a process for generalizing a plan produced by STRIPS so that problem-specific constants appearing in the plan are replaced by problem-independent parameters.The generalized plan, stored in a convenient format called a triangle table, has two important functions. The more obvious function is as a single macro action that can be used by STRIPS—either in whole or in part—during the solution of a subsequent problem. Perhaps less obviously, the generalized plan also plays a central part in the process that monitors the real-world execution of a plan, and allows the robot to react "intelligently" to unexpected consequences of actions.We conclude with a discussion of experiments with the system on several example problems."Artificial Intelligence 3:251-288
For the past several years research on robot problem-solving methods has centered on what may one day be called'simple' plans: linear sequences of actions to be performed by single robots to achieve single goals in static environments. This process of forming new subgoals and new states continues until a state is produced in which the original goal is provable; the sequence of operators producing that state is the desired solution. In the case of a single goal wff, the objective is quite simple: achieve the goal (possibly while minimizing some combination of planning and execution cost). The objective of the system is to achieve the single positive goal (perhaps while minimizing search and execution costs) while avoiding absolutely any state satisfying the negative goal.
This paper proposes a method for handling the frame problem in representing conceptual, or natural-language-type information. The method is part of a larger calculus for expressing conceptual information, called P c F-2, which is described in Sandewall (1972), and which is a modification and extension of Sandewall (1971a). When the STRIPS schema adds a fact, PLANNER would add the corresponding fact to the data base using the primitive thassert. In this context, by epistemological information we mean a notation together with a set of rules (for example, logical axioms) which describe permissible deductions.
We may regard the subject of artificial intelligence as beginning with Turing's article'Computing Machinery and Intelligence' (Turing 1950) and with Shannon's (1950) discussion of how a machine might be programmed to play chess. In this case we have to say that a machine is intelligent if it solves certain classes of problems requiring intelligence in humans, or survives in an intellectually demanding environment. However, we regard the construction of intelligent machines as fact manipulators as being the best bet both for constructing artificial intelligence and understanding natural intelligence. Given this notion of intelligence the following kinds of problems arise in constructing the epistemological part of an artificial intelligence: I.
I shall discuss automatic methods of search for solutions in problems susceptible of a particular formal representation, namely that on which the Graph Traverser program (Doran & Michie 1966, and see Doran p. 105) has been based. One approach, based on state-evaluation, generates all the states of the problem which can be reached in a small number of moves from the current state, and then seeks by some process of evaluation to decide which state shall form the next point of departure. In the classical studies of Newell, Shaw & Simon (1960) selection is applied by going down a priority sequence of operators, applying to each in turn a number of tests, first of applicability to the current state and then of whether the operator conduces towards one or another of various desirable intermediate states, or subgoals.
This article is concerned with the psychology of human thinking. It setsforth a theory to explain how some humans try to solve some simpleformal problems. The research from which the theory emerged is intimatelyrelated to the field of information processing and the construction of intelligentautomata, and the theory is expressed in the form of a computerprogram. The rapid technical advances in the art of programming digitalcomputers to do sophisticated tasks have made such a theory feasible.It is often argued that a careful line must be drawn between the attemptto accomplish with machines the same tasks that humans perform, andthe attempt to simulate the processes humans actually use to accomplishthese tasks. The program discussed in the report, GPS (General ProblemSolver), maximally confuses the two approachesâwith-mutual"!benefit. Lerende Automaten, Munich: Oldenberg KG
The typical pattern-recognition program is either elaborately preprogrammed to process specific arrays of input patterns, or else it has beendesigned as a tabula rasa, with certain abilities to adjust its values, or "learn."One mode of operation of the present program is to begin with no operators at all. In this case operators are initially generated by the program at a fixed rate until some maximum number of operators is reached. The continual replacement of poor operators by new ones then tends to produce an optimum set of operators for processing the given array of inputs. Reprinted in Fegenbaum & Feldman, Computers & Thought (1963).Proc. Western Joint Computer Conference, 19: 555-570.
Even the earliest computers could do arithmetic superbly, but only very recently have they begun to read the written digits that a child recognizes before he learns to add them. Understanding speech and reading print are examples of a basic intellectual skill that can variously be called cognition, abstraction or perception; perhaps the best general term for it is pattern reecognition. Except for their inability to recognize patterns, machines (or, more accurately, the programs that tell machines what to do) have now met most of the classic criteria of intelligence that skeptics have proposed.Scientific American [August, 1960] 203: 60-68. Computers and Thought, Section 6 (1963).
Limited success has been achieved, in punchedcard tests, in improving the idiomatic quality and so the intelligibility of an initially unsatisfactory translation, by word-for-word procedures, from Italian into English, by using a program which permitted selection of final equivalents from "heads" in Roget's Thesaurus, i.e. As a result of discussions which followed, a Research Unit was formed at Cambridge, with the support of the National Science Foundation of the United States, to investigate these problems further.