Fault Detection, Isolation, and Localization in Embedded Control Software

Abstract

Embedded control software reacts to plant and environment conditions in order to enforce a desired functionality, and exhibit hybrid dynamics: control-loops together with switching logic. Control software can contain errors (faults), and fault-tolerance methods must be developed to enhance system safety and reliability. We present an approach for fault detection and isolation that is key to achieving fault-tolerance. Detection approach is hierarchical involving monitoring both the control software, and the controlled-system. The latter is necessary to safeguard against any incompleteness of software level properties. A model of the system being monitored is not required, and further the approach is modular and hence scalable. When fault is detected at the system level, an isolation of a software fault is achieved by using residue methods to rule out any hardware (plant) fault. We also proposed a method to localize a software fault (to those lines of code that contain the fault). The talk will be illustrated through a servo control application.