UCSC-CRL-04-04: Improving Control Performance using Adaptive Quality of Service in a Real-Time System

12/31/2004 09:00 AM
Computer Science
Traditional control systems employ fixed sampling intervals. Recent work in integrated control and realtime systems has resulted in control systems in which the sampling interval may vary based on the state of controller performance. Soft real-time systems provide mechanisms for dynamically adapting application resource usage based on system state and application needs. In this paper we investigate employing this mechanism to allow several control applications to dynamically adapt their resource usage so that they receive enough of the limited resources to achieve their goals, but do not greedily consume resources, allowing the system to be utilized by other applications as well. This paper presents a framework in which a flexible, integrated real-time system directly supports adaptive control applications. Our results show that this technique can result in significantly lower controller error (by an average of over 20% in our experiments) with no increase in overall resource usage.