John Searle's attack on various interpretations of Artificial Intelligence has been the most thorough challenge to the philosophical foundations of Artificial Intelligence. In this paper we attempt to contribute to a growing body of argument as to why Searle is mistaken in his attack. We propose an analogy between intelligent objects and flying objects, leading to a definition of Artificial Intelligence similar to that of aerodynamics, but one which attempts to present general laws of intelligence in man and machines alike.

This paper reviews some current research problems in Artificial Intelligence applied to robotics, in particular the processing of sensory information and robot programming. It is perceived that much progress has been made in applying AI techniques to particular isolated tasks, but the important theme at the leading edge of the AI-based robotics technology seems to be the interfacing of the subsystems into an integrated environment. This requires flexibility of the subsystems themselves. But on the other hand it also increases the flexibility of the integrated system in that it broadens the variety of ways for solving problems through functional combination of the subsystems. The presentation in this paper is illustrated by examples from the hardware and software environment of the advanced robotics research at the Turing Institute.

This paper describes the principles of the advanced programming techniques often dubbed Artificial Intelligence involved in decision making as may be of some value in matters related to production engineering. Automated decision making in the context of production can adopt many aspects. At the most obvious level, a robot may have to plan a sequence of actions on the basis of signals obtained from changing conditions in its environment. These signals may, indeed, be quite complex, for example the input of visual information from a television camera.

At another level, automated planning may be required to schedule the entire work cycle of a plant that includes many robots as well as other types of automated machinery. The often-quoted dark factory is an example of this, where not only some of the operations (such as welding) are done by robots, but also the transport of part-completed assemblies is automatically scheduled as a set of actions for autonomic transporters and cranes. It is common practice for this activity to be preprogrammed to the greatest detail. Automated decision making is aimed at adding flexibility to the process so that it can absolve the system designer from having to forsee every eventuality at the design stage.

Frequent reference is made in this context to artificial intelligence (AI), knowledge-based and expert systems. Although these topics are more readily associated with computer science, it is the automated factory, in general, and the robot, in particular, that will benefit from success in these fields. In this part of the paper we try to sharpen up this perspective, while in part II we aim to discuss the history of artificial intelligence in this context. In part III we discuss the industrial prospects for the field.

Artificial Intelligence clearly influences Robotics, but the advent of the latter alters the character of A.I. itself, bringing it closer to natural intelligence. This is partly due to greater attention to processes depending on continuous variables, and the combination of these with concept-based or “logical” processes. Some fundamental A.I. principles, notably Minsky's heuristic connection, involve continuity, and the advent ofRobotics should stimulate developments which take them into account. A scheme for a robot which can increase its speed of operation by a learning process is outlined.

A method for solving the recognition of partially occluded parts is presented. It is based on the automatic generation of features from a set of primitive features which are configurations of pairs of fixed length segments of boundary edges of the parts. The procedure that creates the recognition features assigns a number in the range (0,1) that indicates the importance of the feature in the recognition strategy. This number is referred to as the feature's saliency. The method assumes that the parts that can occur in a scene come from a known set of parts. An example illustrates how automatically generated features can be used to count the number of identical parts in a heap.

We suggest a new pretopological model for pattern recognition which was introduced to study complex economic systems. The model has its origin in the concept of “neighbour”, which is both primitive and fundamental in pattern recognition. Pretopology enables us to develop a perceptive and topological approach for patterns and to see that problems, apparently different, are in fact identical e.g. clustering and recognition, search of skeletons in image processing and search of an informative learning set. It should be noted that the suggeted model is more than a descriptive one.