Resource and performance trade-offs in real-time embedded control systems

摘要

Embedded control systems are subject to conflicting demands: end-users ask for devices with better capabilities while strong industrial competition impose tight cost constrains that results in devices with limited resources. Current research in the multidisciplinary embedded systems discipline indicates that by combining realtime and control systems it is theoretically feasible to design resource-constrained embedded control systems capable of trading-off control performance and resource usage. This paper focuses on the implementation feasibility of recent state-of-the-art resource/performance-aware (RPA) policies that can be applied to a set of control loops that concurrently execute on a microprocessor. The objective of these policies is to improve control performance and/or to minimize resource utilization. The paper first reviews existing state-of-the-art RPA policies. Then it presents a performance evaluation framework (PEF) that permits to assess whether RPA policies can be implemented in practice. The PEF is designed using a modular approach and following the guidelines obtained by a taxonomic analysis performed on the state-of-the-art RPA policies. Finally, a case study is presented where the PEF is applied to a set of representative RPA policies. The case study reveals that the modularity of the PEF allows tailoring the framework to evaluate any specific RPA policy, which indicates that RPA policies can be implemented in practice. But it also reveals that the problem of assessing diverse RPA policies in fair conditions implies facing and solving conflicting demands by even taking decisions that may not favor equal all policies under evaluation. Nevertheless, the comparative analysis permits identifying potential benefits and drawbacks of each policy, as well as extracting design guidelines for future real-time embedded control systems theory and practice.