An intelligent control system using object model by real-time learning

Safety and reliability improvement is an important for an automated system. One of the approaches to improve these is that preventing accidents even if the system malfunctions. This is difficult to ensure when using conventional control mechanisms, as these mechanisms may not be applicable if the controlled object has been changed by the malfunction. On the other hand, humans are able to adjust and act flexibly in response to a changing situation. In this paper, an intelligent control system is proposed that emulates this human capacity to adjust to a changed situation. The key feature of this control system are "hierarchical mechanism" and "acquisition of a new control rule and control according to the rule". To acquisition of a new control rule, the predictive model in real time learning the object is build into this control system. This control system is applied to a flight control system, and its effectiveness for the improvement of safety and the reliability is confirmed.