Supervisor: Andreas Steininger
Towards Self-Healing Asynchronous Circuits Abstract
Faults were mainly considered a threat in applications operating in remote environments or have long-term missions where, in case of a malfunction, traditional repair by human operators is often very expensive or not even possible, such as in space missions. Nowadays, due to advancing miniaturization and continuously decreasing supply voltages, even non-safety-critical systems are more susceptible to faults. This becomes a serious problem for high reliability applications. The need for critical devices to keep on working even with some loss of performance, leads to a demand for runtime permanent-fault-tolerant systems. The traditional fault tolerance mechanisms are usually very costly in terms of resources. Even though there are serious consequences to fault occurrences, most research targets transient faults, and there are hardly any concepts available to deal with the permanent ones. QDI circuits o er the attractive fail-stop behavior in the presence of permanent faults. The compelling properties possessed by asynchronous logic, especially its inherent robustness against faults, makes it a solution worth exploring.
In this thesis, we aim to examine and quantify the relative occurrence rate for each e ect manifested in the circuit by a permanent fault. We will use formal description methods for asynchronous circuits to model the faults in targeted circuits, as well as simulations. As a second step, we plan on using the statistics, obtained from our earlier investigations, to improve the circuit’s fault tolerance and to turn it into an adaptive, self-healing system that is capable of autonomously coping with failure situations.
We plan on verifying the e ectiveness of our concepts through experiments, both simulation and im-plementation, on an FPGA ( eld-programmable gate array) with extensive fault injection tests.