Despite many years of research in the real-time systems research community, no practically useful probabilistic timing analysis exists today. In this project we aim to overcome this situation by addressing key challenges inherent in the restricting assumptions of existing analysis. These challenges will be approached from a new direction compared to what has been tried before, resolving complicating circumstances inherent in dependencies among system components. Novel run-time mechanisms will be developed to better support both probabilistic timing and performance guarantees.
The project is organized in 3 work packages targeting component- and system-level modeling, related analysis, and runtime mechanisms, respectively. Solutions will be developed, integrated and evaluated, both in the laboratory and in more realistic settings together with industrial partners.
The key motivation for probabilistic analysis is that it provides means for a more cost conscious and balanced allocation of system resources. Since other system components may fail with a higher probability than the processing resources - why design the processing resources for a scenario that is unlikely to occur? The proposed new run-time mechanisms and probabilistic timing analysis techniques have, by providing sufficient level of guarantees without overprovisioning resources, a potential to significantly advance probabilistic real-time systems research as well as adoption of real-time systems research in industry.
|First Name||Last Name||Title|
Change-point and model estimation with heteroskedastic noise and unknown model structure (Jul 2023) Anas Al-hashimi, Thomas Nolte, Alessandro Papadopoulos 9th International Conference on Control, Decision and Information Technologies (CoDIT 2023)
Continuous-Emission Markov Models for Real-Time Applications: Bounding Deadline Miss Probabilities (May 2023) Anna Friebe, Filip Markovic, Alessandro Papadopoulos, Thomas Nolte 29th IEEE Real-Time and Embedded Technology and Applications Symposium (RTAS 2023)
Analytical Approximations in Probabilistic Analysis of Real-Time Systems (Dec 2022) Filip Markovic, Thomas Nolte, Alessandro Papadopoulos 43rd IEEE Real-Time Systems Symposium 2022 (RTSS2022)
Adaptive Runtime Estimate of Task Execution Times using Bayesian Modeling (Aug 2021) Anna Friebe, Filip Markovic, Alessandro Papadopoulos, Thomas Nolte 27th IEEE International Conference on Embedded and Real-Time Computing Systems and Applications (RTCSA'21)
On the Convolution Efficiency for Probabilistic Analysis of Real-Time Systems (Jul 2021) Filip Markovic, Alessandro Papadopoulos, Thomas Nolte 33rd Euromicro Conference on Real-Time Systems (ECRTS 2021)