Collaborative work between human and machine during an accident in nuclear power plant

A nuclear power plant (NPP) is a complex system. Traditionally, it is considered that recovery from an accident can be made by experienced and well-trained humans rather than by a computerized support system. However, recent advanced information technology enables us to study and to discuss the new machine systems that support the human operators. Recent studies on human-machine interaction have recommended a human-machine collaborative system rather than a full automated one. One kind of human-machine collaboration is “human-centered automation”. In the context of human-machine interaction, we introduce the concept of an “agent” into the design of a machine system to support human operators. The agent system proposed in this paper has its own situation awareness and manages two resources: safety resources and time resources. The agent can generate adequate operations for accident recovery, taking the situation and the resources into account. The agent switches the collaborative modes and the target according to the situation. This switching function of the collaborative modes allows the agent system to support the operator as a human supervisor does. In this paper, we treat the architecture of the agent system under the following headings: (1) collaborative modes and the switching function, (2) the description of the situation, (3) a method for the estimation of available time, and (4) the human-machine interface. Additionally, this paper concerns a study of an agent system that can support human operation teams as an extension of the aforementioned study