Intent Inference with Path Prediction

Abstract

In today's air-traffic management system, the intent of an aircraft is revealed in its flight plan and a pilot's adherence to published navigation routes. In a future free-flight environment, aircraft might be allowed to fly any route they choose. Intent will likely be broadcast in the form of trajectory change points (TCPs), for instance, up to four TCPs in an automatic dependent surveillance-broadcast message. However, if broadcast TCPs do not accurately represent intent, do not exist, or do not get received, then a nearby aircraft or ground-monitoring system has a need to infer the pilot's intent in real time. In this paper, a method of inferring intent, which is based on artificial intelligence models and a process for best fitting an intent model to observed aircraft motion, is investigated. Horizontal, vertical, and speed dimensions are first investigated independently, and then combined, including sequences of actions, to fully explain the three-dimensional guidance and navigation plan of an aircraft. Finally, the inferred intent is used as a basis to predict a path from the current location of the aircraft into the future.