As the performance of computers improves, the use of digital controllers has been increased in a wide variety of fields. The design procedure for real-time control system is composed of two stages. One is the controller design stage, and the other is the implementation stage.
In the controller design stage, various digital control theories are utilized with assumptions: synchronous sampling, no sampling jitter, and negligible feedback latency (latency from sensing to actuation). However, in the implementation stage, multiple control tasks are usually scheduled on a processor, which causes finite sampling period, varying feedback latency, and sampling jitter, and hence the assumptions in the controller design stage are no longer valid. These scheduling-induced delay and jitter degrade the control performance. In order to obtain better control performance, it is necessary to consider the effect of real-time scheduler on digital controllers.
In this dissertation, we investigate the relationship between control performance and timing parameters such as sampling period, feedback latency, and jitter. Using this relationship, we formulate a new performance index with timing parameters. The new performance index with timing parameters can help a system designer to make a trade-off in real-time scheduling. In order to measure the effects of timing parameters, we define the degradation factor of each timing parameter, and found that the degradation factor of feedback latency is larger than those of other timing parameters.
Using these property, we propose a heuristic algorithm, near-optimal period assignment algorithm, for maximizing control performance. We show that proposed algorithm finds optimal period set in feasible area much faster than full search method. In addition, we introduce scheduling algorithms which improve overall control performance under limited computing resource condition.
First, we propose non-preemptive last section(NPLS) assignment algorithm whic...