Mitra Nasri did her PhD thesis on analyzing and improving quality of service of real-time embedded systems in dynamic environments. Her thesis is mostly concerned about improving quality of service and accuracy of real-time systems, where the accuracy is defined as the closeness of system state to the goal state. In many industrial real-time embedded systems working in highly dynamic environments, system safety which depends on guaranteeing deadlines may not be sufficient; rather, to improve system’s quality of service (QoS), the instants of input and output operations are of high importance to deal with the rapid changes of the environment. For example, in control systems, measures such as sampling delay (delays before input), I/O delay (delays between input and output operations), their jitters, and quality of data considerably affect their accuracy.
To improve the accuracy, she has introduced two groups of scheduling algorithms; preemptive and non-preemptive. In the first group, the goal is to increase system accuracy by modifying priorities of the tasks at run-time. In this group she has developed several scheduling algorithms with different properties to reduce delays (sampling and I/O delays) and to limit jitters to user defined values. For most of the algorithms in this group, she has provided schedulability tests to verify sufficient (and in some cases, necessary) conditions of schedulability of hard real-time systems. In the second group, she introduced non-work-conserving scheduling algorithms with guaranteed schedulability in special cases of harmonic tasks. For some of those algorithms, she has provided sufficient conditions for a jitter free schedule. Finally, she has developed a framework to construct customized harmonic periods from given period ranges. Using this framework, feasible task sets for the developed algorithms can be obtained.
Morteza Mohaqeqi did his PhD thesis on probabilistic thermal analysis and management of embedded real-time systems. Inter effects between performance parameters, dynamic and leakage power consumption, and temperature variation were considered. For thermal modeling and analysis of stochastic real-time systems, an analytical approach was adopted. To this aim, on the basis of a Markovian model of a stochastic real-time system, steady-state thermal characteristics of the processor was derived, which is used for power and reliability calculation. For thermal management of the real-time system, an optimal control approach, namely model-predictive control (MPC) was employed. In that work, the controller determines the speed of processor and cooling system to minimize the average power consumption, while respecting the performance constrains. Also, the existing theory of stochastic control was used for thermal management of a multicore real-time system. The goal of that work was controlling the temperature gradient of the processor through migration mechanism.