The standard platform architecture for real-time embedded systems has increasingly shifted away from single processor platforms to multiprocessor platforms. The recent shift towards multiprocessor platform architectures has resulted in increased consolidation and integration of multiple subsystems upon shared processing platforms, aided in part by virtualization execution environment (VEE) technologies. The potential impact of the subsystem-integration approach is a significant reduction in the size, weight, and power (SWaP) requirements of integrated systems over non-integrated systems. However, the tighter physical integration of subsystems upon a shared processing platform introduces fundamental questions on how the processor's computational resources should be effectively allocated among the contingent subsystems and how temporal isolation between real-time subsystems should be achieved.
The overall objective of this NSF CAREER project is to obtain solutions to the above questions via development of effective real-time scheduling algorithms, formal analysis, and tools for supporting tunable temporal isolation of subsystems upon a multicore VEE platform. The specific research objectives of the project are: effective system and subsystem real-time scheduling algorithms for VEE frameworks, protocols for resource sharing between subsystems, schedulability analysis, and implementation of a VEE for controlling a real-time robotic system. The educational goal of the project is to increase overall awareness and understanding of the importance of building and verifying temporally correct systems by: recruiting graduate and undergraduate students from underrepresented populations into embedded systems research; developing embedded systems curriculum and outreach training program; introducing K-12 students to general embedded systems concepts; and developing online real-time systems repository for students, researchers, and faculty.
