Runtime Embedded Software Malfunction Detection Based on Profile Generation of Current-Level Pattern Matcher

As the demand for high-performance embedded systems grows, the complexity of embedded software increases, potentially leading to software errors and system malfunctions. In this paper, we propose a system to monitor malfunctions caused by software errors by extracting the normal operating current patterns of a target embedded system (TES) in actual operation and comparing them with the real time current TES data. By using a monitoring embedded system (MES), the normal operating current patterns can be automatically generated, making it easy to promptly detect malfunctions in the exposed TES. When the MES detects a malfunction, it applies algorithms to efficiently recover the malfunctioning TES. The proposed system employs two algorithms when a malfunction in the TES is detected. First, upon detecting a malfunction, the MES resets TES to restore it to normal operation. If the malfunction persists after the reset, the MES controls the TES to completely halt operation. Second, if a malfunction is detected, it stabilizes the abnormal current state to the normal current state through a proportional integral derivation (PID) control. The TES typically consumes about 75 to 95 mA of current during normal operation. The MES applies a specific error detection rate of 20 %, considering any consumption above 114 mA a malfunction and using the algorithms to control it. The system has been verified to reset TES upon detecting a malfunction and to stabilize the operational current through PID control during abnormal current states.