Abstract Performance optimization of cyber-physical systems （CPS） calls for co-design strategies that handle the issues in both computing domain and physical domain. Periods of controller tasks integrated into a uni...Abstract Performance optimization of cyber-physical systems （CPS） calls for co-design strategies that handle the issues in both computing domain and physical domain. Periods of controller tasks integrated into a uniprocessor system are related to both control performance and real-time schedu- lability analysis simultaneously. System performance improvement can be achieved by optimizing the periods of controller tasks. This paper extends an existing model to select task periods in real-time for CPS with fixed priority controller tasks scheduled by rate-monotonic algorithm. When all the tasks can be integrated, the analytic solution of the problem is derived by using the method of Lagrange multipliers and gradient descent method is evaluated to be suitable online. To further deal with the condition that the system is overloaded, an integrated method is proposed to select periods of tasks online by selecting a subset of tasks first and then optimizing the periods for them. Experimental results demonstrate that our method yields near-optimal result with a short running time.展开更多
基金supported by State Administration of Science,Technology and Industry for National Defense,China(No.1000-GEAC0001)
文摘Abstract Performance optimization of cyber-physical systems （CPS） calls for co-design strategies that handle the issues in both computing domain and physical domain. Periods of controller tasks integrated into a uniprocessor system are related to both control performance and real-time schedu- lability analysis simultaneously. System performance improvement can be achieved by optimizing the periods of controller tasks. This paper extends an existing model to select task periods in real-time for CPS with fixed priority controller tasks scheduled by rate-monotonic algorithm. When all the tasks can be integrated, the analytic solution of the problem is derived by using the method of Lagrange multipliers and gradient descent method is evaluated to be suitable online. To further deal with the condition that the system is overloaded, an integrated method is proposed to select periods of tasks online by selecting a subset of tasks first and then optimizing the periods for them. Experimental results demonstrate that our method yields near-optimal result with a short running time.
