Analysis of real-time multi-modal FP-scheduled systems with non-preemptible regions

Abstract

Over the years, multiple hardware and software operating modes have been employed in many computing devices (e.g., tablets, smart-phones, GPS receivers) to efficiently utilize device resources. Similar advantages are also preferred in realtime systems (RTS) due to the requirement that a RTS must respond in a timely manner to a physical environment that may change sporadically. An efficient multi-modal system (MMS) is also a prerequisite for the development of real-time control systems which can maintain stable system behavior while ensuring timing guarantees for a changing set of real-time tasks. However, the currently-available fixed-priority (FP) schedulability analysis for multi-modal systems with both software/hardware modes is computationally expensive. In addition, current schedulability analysis for systems that support mode changes requires an assumption that is often not suitable for cyber-physical systems (CPS): sensing and actuation in the underlying physical plant are preemptible activities. However, sensors such as radar transmitter/ receiver requires non-preemptible access to the processor upon sending and then processing the return signal for accuracy. In this research, we develop a framework for multi-modal RTS scheduled by FP algorithm along with efficient schedulability analysis with pseudo-polynomial complexity considering the advantages and limitations of specific software/hardware model. Two simulations: a case study on adaptive cruise control in automotive systems, and schedulability comparison are included to corroborate the performance of the schedulability analysis.