RÉSUMÉ Ce texte présente d’abord les différentes situations d’ordonnancement et les questions qu’elles posent, tant aux machines qu’aux humains. Il fait ensuite le point sur les travaux menés depuis la dernière revue de question exhaustive de P. Sanderson (1989) sur les facteurs humains dans l’ordonnancement. Il rappelle l’état de la question en 1989 et décrit quelques avancées dans le renouvellement des approches formelles, dans l’analyse des stratégies humaines et dans l’émergence d’une conception plus coopérative des relations homme-machine dans ce domaine. Il se conclut par quelques directions de recherche qui pourraient être prometteuses.
HUMAN OPERATOR SUPPORT OF SCHEDULING IN MANUFACTURING WORKSHOPS This paper presents three scheduling situations from a technical point of view. First, static scheduling is considered as the design of a schedule before its execution. Second, dynamic scheduling considers the elements of the problem in real time, as they are known. Third, reactive scheduling consists of modifying a schedule on the basis of unexpected events. A number of questions arise from these situations ; for humans in the main, but also for machines. How do humans reduce complexity ? How do they manage uncertainty ? What is the nature of human expertise ? What is the human role in the scheduling process ? What degree of confidence should the human have in the machine’s model of the problem ? Is the human able to improve on the solution provided by the machine ? What kind of diagnosis is necessary when deciding and implementing a modification in a schedule ? The paper then looks for responses to these questions in the literature. The state-of-the-art work on human factors in scheduling, published by P. Sanderson in 1989, is considered. Three main contributions to research were current at this time. First, a comparison between human strategies and simple scheduling rules showed the frequent superiority of the former over the latter. Second, hybrid systems (human-machine cooperation) were designed and evaluated as being better than humans or machines alone. Third, human-computer interfaces were evaluated, integrating prediction and graphical features. A limitation of these studies was their poor ecological validity, as they were restricted to laboratory settings. The paper then describes the progress made in the revival of these approaches. Through individual protocol analysis, a more precise access to human strategies was possible. Formal approaches were enriched with contributions from distributed cognition. Some progress has also been made in the analysis of human-computer interfaces developed within the context of ecological interfaces. Finally, a cooperative conception of the human-machine relationship has grown within this domain, as well as a desire to support human machine cooperation in distributed scheduling. The paper concludes with some promising research directions that could result in a better understanding of human strategies and representations, the design of more efficient human-machine interfaces and cooperation, and a greater consideration of dynamic and reactive scheduling.
