Adaptive fuzzy controller for predicting order of concurrent actions in real-time systems

In controlling a real-time system with a large number of controlling points and when the time spent for setting up the reacting actions is longer than the deadline, we have to predict the reacting actions before receiving complete information about them. These actions are chosen from the concurrent past actions based on the rank of their utilities. Thus in predicting the rank we have to estimate the indefinite qualities defined by the uncompleted information. The paper presents a hardware approach to predict the order of concurrent actions based on a fuzzy and adaptive mechanism which measures the possible utilities given in intervals with additional statistical information. In order to process efficiently the incomplete data we design a fuzzy hardware controller (FHC) which specializes on ranking the intervals according to the observed data in the past. The analysis shows that the fuzzy measurement implemented by a hardware specialized controller with parallel architecture is able to reduce significantly the processing time, compared with implementation by software running on a general computer. Moreover, the FHC can adapt to the types of statistical data which is also safer and can be updated easily within the common memory of the FHC.