Real Time Heading Sensors Fusion and Fault Detection

In modern offshore racing, performance often depends on two main factors: a good autopilot and the right strategy decisions taken by the skipper. Some sensors are crucial to ensure the quality of those two keys of success, among which we can mention the heading sensors. Unfortunately, those sensors, whether magnetometers or GNSS based, are subject to disturbances and faults of various origins: magnetic disturbances from other devices, GPS fix or reception issues, sensor drift, etc. These sensor faults can cause an autopilot’s solution to diverge which can result in serious damage to the boat or the crew. Assurance of a valid measurement is therefore a key point to ensure reliability of autopilot systems and skipper’s decisions. This paper presents a method to produce consistent values of true heading and yaw rate while detecting sensor faults. The proposed solution relies on the hypothesis that sensors using different technologies and placed in different spots inside the boat will not be subject to identical and synchronised disturbances. Thus, by intelligently fusing the information coming from several sources, a continuous and consistent true heading measure can be maintained. A simple dynamic model for the heading and yaw rate is implemented and an asynchronous filter update is done depending on available measures. The difference between the estimated and the measured states is used to determine whether a sensor is faulty or valid and the update is done consequently; then the information on sensors status and quality of the estimation can be propagated. In the paper, we detail a method to detect faults in heading sensors and to provide a substitution value if necessary. The proposed model is validated by test campaigns that were conducted using both data logs and on-board tests. Results show that we can improve and maintain true heading measurement quality and detect and isolate faulty sensors.