Man-machine supervision

The complexity of the systems in which man is involved today is leading to the emergence of increasingly sophisticated and inescapable information processing systems, where decision-making is becoming more and more difficult. Man-machine supervision of production systems is a typical example of this type of problem.

Dictionary definitions of supervision are: "to supervise and control the execution of an operation or the performance of work by others" (Larousse); "to control without going into details" (Robert). The two notions of "supervising and controlling the execution of an operation" and "without going into details" are key to understanding the orientations described in this article.

The operator's role has evolved from that of driver to that of supervisor, and the production tool has become inseparable from its digital control system, making understanding events all the more complex. The integration of an ever-increasing number of sophisticated control indicators in the operator station does not necessarily meet the expectations of the operating team. It is preferable to develop cooperative information systems, which are genuine aids to reasoning and understanding situations for operators. The aim should not be to automate operators' cognitive tasks, but to help them reason. Supervision cannot be reduced to the monitoring of a physical process; the object of its analysis is a complete installation, with its instrumentation, operating modes, configurations, etc., run by a team of operating operators.

One of the challenges of interactive information systems then lies in the selection, organization and dynamic presentation of information; the performance of the man-machine assembly depends on the effectiveness of the communication established by the interface (in the broadest sense of the term). The design of cooperative systems relies on the analysis of the man-machine system to establish information needs, define objectives, constraints and tasks to be fulfilled.

This article describes the principles and specific features of human-machine supervision. We will contrast the notion of cognitive automation with that of reasoning aid. This will lead us to discuss cognitive models useful to the operator in the control room. We will focus on qualitative reasoning, causal reasoning, multi-point of view reasoning and approximate reasoning.