M-Group – Extending the Renode-based Experiment Platform

Within M-Group there are multiple research tracks. One research track investigates how embedded software can be made more resilient against hardware bitflips through software- implemented fault tolerance. This track has resulted in multiple fault detection techniques, a compiler extension and a fault injection tool that can test the fault detection techniques.
Another research track investigates how machines can be retrofitted to detect anomalies through machine learning on the edge. This track has resulted in a new and improved anomaly detection technique that can be generalized over many machine conditions.

In first studies, these two tracks have been brought together, through two machine learning case studies executing on an nRF52840 DK board and through their simulated variant provided by the Renode framework. This thesis will build upon this work by further extending the Renode-based experiment framework.

• Optimize the fault injection by using multiple simultated boards and dividing the experiment among them.
• Using a different platform, in order to generalize the connection/interaction interface

And if time allows: investigate if an I/O driven demonstrator of the M-Group, i.e. our small scale factory, can be fully simulated and experimented through Renode