Research on energy-aware real-time control system design with dynamic voltage scaling

Abstract

Dynamic voltage scaling (DVS) technique has become a widely-studied method for building energy efflcient real-time embedded systems, however, little effort has been made to use them for building energy-aware control system. In this dissertation, various approaches are made to apply DVS techinique in design of control systems. A problem that may occur while trying to schedule conventional control task on a DVS scheduler is introduced. Previous performance index used in control system has been extended to include both control performance and processor energy consumption to measure overall performance of proposed design methods. First approach is to assign a task period for each control task to be scheduled on a conventional DVS scheduler that results in best overall performance for given environmental constraints. This method is an offline heuristic method which is much faster in finding solution than full search simulation method. Second approach also considers use of conventional DVS scheduler, but seeks for online performance enhancement through use of adaptive periods for scheduled control tasks. Third approach is different from previous two in that new online scheduling algorithm is proposed that schedules processor voltage as well as control task period at the same time. Proposed methods are verified through software simulation using MATLAB with True- Time simulator and through experiment on actual hardware using proposed DVS Control System middleware.