Software-based fault detection and recovery for cyber-physical systems

Abstract

Cyber-physical systems demand higher levels of reliability for several reasons. First, unlike traditional computer-based systems, cyber-physical systems are more vulnerable to various faults since they operate under harsh working conditions. For instance, sensors and actuator may not always obey their specification due to wear-out or radiation. Second, even a minor fault in cyber-physical systems may lead to serious consequences since they operate under minimal supervision of human operators. In this paper we propose a software framework of fault detection and recovery for cyber-physical systems, called Fault Detection and Recovery for CPS (FDR-CPS). FDR-CPS focuses on specific types of faults related to sensors and actuators, which seem to be the likely cause of critical system failures such as system hangs and crashes. We divide such critical failures into four classes and then present the design and implementation of FDR-CPS that can successfully handle the four classes of critical failures. We also describe a case study with quadrotor to demonstrate how FDR-CPS can be applied in a real world application.