Self-Healing Spatio-temporal Data Streams Using Error Signatures

Self-healing spatio-temporal data streaming systems enable error detection and data correction based on error signatures. Error signatures are mathematical function patterns with constraints and are used to identify and categorize errors in redundant spatio-temporal data streams. In this paper, we apply these methods to real data from a private Cessna flight and from the Air France AF447 accident in June 2009. For the private Cessna flight, three error scenarios are simulated: pitot tube failure, GPS failure, and simultaneous pitot tube and GPS failures. The error detection accuracy is approximately 93% and the response time to correct data is at most 5 seconds. For the AF447 flight, 162 seconds of available flight data including the pitot tubes failure is collected from the accident report. The pitot tube failure of the AF447 flight is successfully detected and corrected after 5 seconds from the beginning of the failure. Overall error mode detection accuracy reaches 96.31%.Furthermore, our simulations show that the system never corrects data incorrectly, i.e., all inaccurate mode detections produce either unknown or unrecoverable errors.