Section: New Results

Probabilistic Solutions for Hard Real-Time Systems

Participants : Adriana Gogonel, Dorin Maxim, Antoine Bertout, Tomasz Kloda, Irina Asavoae, Mihail Asavoae, Cristian Maxim, Walid Talaboulma, Slim Ben-Amor, Robert Davis, Liliana Cucu.

The probabilistic solutions for hard real-time systems are built under the hypothesis that worst case values and worst case execution scenarios have extremely low probability of appearance. While continuing the estimation of bounds for the worst case execution times of a program [34], [25], we have proposed the first utilisation of probabilistic description for mixed-criticality systems [42]. Our result is exploiting the heavy tails of the execution times of a program to propose efficient scheduling solutions. Moreover since the feasibility intervals [21] for a probabilistic real-time system is not formally identified, we have formulated the first feasibility reasoning for such systems [47] under fixed-priority assignment policies [20]. Another important problem for probabilistic real-time systems concerns the feasibility in presence of precedence constraints, often used by our industry partners. The introduction of precedence constraints requires the comparison of probabilistic arrivals and we showed that existing measures are not correct in this context and we proposed and proved correct new measures [24].