Experiments illustrating the importance of automated reasoning

Experimental results that demonstrate the need to incorporate automated reasoning in the design of a controller for a spacecraft nuclear reactor are presented. It is argued that control systems intended for safety-constrained nonlinear processes should not only provide such traditional functions as tracking a demanded trajectory but also incorporate methods for planning, prediction, and assessment, without which the automated diagnosis of malfunctions and hence the online reconfiguration of equipment may not be possible. Experimental results are described that show the relevance of initial plant conditions to the proper planning of a maneuver, the relation between a controller's capability to halt a transient and the prediction of system states from which it can restore equilibrium conditions, and the role of signal validation and online diagnostics in the assessment of controlled performance. The results were obtained while evaluating period-generated control for the adjustment of reactor neutronic power. The period-generated control approach and reactor dynamics are summarized. >