Context-aware Intelligent Assistant for Decreasing Pilot Workload

Abstract

Although weather-related accidents accounted for only 3.6% of all general aviation accidents in 2002, over 70% were fatal. Faulty decision making due to inaccurate or incomplete awareness of the weather situation accounts for approximately 14% of all fatal general aviation accidents. We describe a context-aware, domain and task knowledgeable, personalized and adaptive assistant designed to decrease the workload required to maintain situational awareness. The assistant automatically monitors weather reports for the pilot’s route of flight and warns of detected anomalies. When and how warnings are issued is determined by phase of flight, the pilot’s definition of acceptable weather conditions, and the pilot’s preferences for automatic notification. In addition to automatic warnings, the pilot is able to verbally query for weather and airport information. By noting what requests are made during the approach phase of flight, our system learns to provide the information without explicit requests on subsequent flights with similar conditions. We show that our weather assistant decreases the time required to answer relevant pre-flight questions by more than 2.5 times over answering the same questions using a conventional pre-flight weather briefing, and decreases the time required to maintain in-flight weather situational awareness by more than 5.5 times when compared to the conventional method of in-flight weather briefings. In both situations, our system enables the pilot to allocate more time to other tasks such as scanning for anomalies.