Multi Sensor Data Fusion for Sensor Failure Detection and Health Monitoring

Abstract

The German Aerospace Center (DLR) Institute of Flight Systems has developed a demonstrator for enhanced autonomous technologies. Safe operation in civilian airspace of the small scale autonomous helicopter is a long term research aim. The use of low cost micro sensors in ARTIS (Autonomous Rotorcraft Test bed for Intelligent Systems) and the high level of noise make intelligent data fusion a mandatory requirement for health monitoring. In this paper model based magnetometer failure detection for ARTIS is implemented and investigated. Commonly encountered sensor failures are emulated using simulated and real magnetometer data. Failure detection is achieved by residual monitoring. The health monitoring utilizes the model based redundancies present in the established flight path reconstruction model, which is based on two sequential Extended Kalman filter. The results indicate the feasibility of the methodology for this type of rotorcraft vehicle. The work suggests that fusion of additional vision sensor data would not only provide an excellent reconfiguration strategy but also enhance the detection of small magnetometer drift failure types.