Abstract:

A method of finite-state machine (FSM) based fault detection is proposed to detect faults of event-driven programs. To describe expectation of program behaviors, users specify FSM specifications and insert relevant FSM annotations into the programs. At runtime, software faults are detected through comparing the FSM specifications with collected execution information of the relevant FSMs. Storage and runtime overheads of three detection mechanisms are investigated on two typical node hardware platforms. Two cases are given to demonstrate the effectiveness of the proposed method for sensor node software fault detection.

Key words: sensor networks, event-driven model, fault detection, finite-state machine