US-12626604-B2 - Real-time aircraft flight delay prediction

US12626604B2US 12626604 B2US12626604 B2US 12626604B2US-12626604-B2

Abstract

Examples are disclosed that related to providing flight-delay estimations of airborne flights in a real-time environment. In one example, an aircraft information message for a current aircraft flight is received. The aircraft information message has a designated format consumable by a machine learning model previously trained to assess delay predictions for aircraft flights. The aircraft information message includes one or more aircraft flight-plan parameters, one or more aircraft surveillance parameters, and one or more weather parameters for the current aircraft flight. The aircraft information message is provided as input to the machine learning model to assess a real-time delay prediction for the current aircraft flight based at least on the one or more flight-plan parameters, the one or more aircraft surveillance parameters, and the one or more weather parameters included in the aircraft information message.

Inventors

Andres Munoz Hernandez
Manuel Polaina Morales
Alejandro Güemes Jiménez

Assignees

THE BOEING COMPANY

Dates

Publication Date: 20260512
Application Date: 20220519

Claims (20)

1 . A computing system comprising: a plurality of computing nodes, each computing node of the plurality of computing nodes including a logic processor and a non-volatile storage device holding instructions executable by the logic processor, wherein the plurality of computing nodes includes a sub-set of two or more prediction computing nodes, each prediction computing node of the sub-set of prediction computing nodes being configured to: receive an aircraft information message for a respective current aircraft flight having a designated format consumable by a machine learning model previously trained to assess delay predictions for aircraft flights, the aircraft information message including one or more aircraft flight-plan parameters, one or more aircraft surveillance parameters, and one or more weather parameters for the respective current aircraft flight, provide the aircraft information message as input to the machine learning model to assess a real-time delay prediction for the respective current aircraft flight based at least on the one or more aircraft flight-plan parameters, the one or more aircraft surveillance parameters, and the one or more weather parameters included in the aircraft information message, wherein the real-time delay prediction for the respective current aircraft flight indicates a predicted arrival time of the aircraft and an indication whether the aircraft is ahead of schedule or behind schedule relative to an expected arrival time, wherein the machine learning model is previously trained with training data including training flight-plan parameters, training aircraft surveillance parameters, and training weather parameters from a plurality of previously-completed aircraft flights, wherein the training data further includes actual arrival times for the plurality of previously-completed aircraft flights and delay predictions for the plurality of previously-completed aircraft flights that include delay predictions for previously-completed aircraft flights that arrived ahead of schedule relative to an expected arrival time and delay predictions for previously-completed aircraft flights that arrived behind schedule relative to an expected arrival time, and wherein the machine learning model is previously trained using weighted training in which a delay prediction having a smaller error relative to a corresponding actual delay is weighted more than a delay prediction having a larger error relative to a corresponding actual delay, and output the real-time delay prediction for the respective current aircraft flight in a format that is consumable by a distributed event streaming platform that is configured to publish data feeds of real-time delay predictions output by the sub-set of prediction computing nodes, including the real-time delay prediction for the respective current aircraft flight output by each computing node of the sub-set of prediction computing nodes; and wherein the plurality of computing nodes includes a computing node that is configured to: receive, via the distributed event streaming platform, a plurality of data feeds of real-time delay predictions for a plurality of current aircraft flights, including the real-time delay prediction for the respective current aircraft flight output by each computing node of the sub-set of prediction computing nodes; extract a plurality of real-time delay predictions for the plurality of current aircraft flights from the plurality of data feeds, including the real-time delay prediction for the respective current aircraft flight output by each computing node of the sub-set of prediction computing nodes; and visually present the plurality of real-time delay predictions for the plurality of current aircraft flights in a graphical user interface (GUI).
2 . The computing system of claim 1 , wherein one or more computing nodes of the plurality of computing nodes is a data consumption computing node configured to: receive a plurality of different data feeds including an aircraft flight-plan data feed, an aircraft surveillance data feed, and a weather data feed, extract one or more aircraft flight-plan parameters, one or more aircraft surveillance parameters, and one or more weather parameters from the plurality of different data feeds for the respective current aircraft flight, and generate the aircraft information message for the respective current aircraft flight, according to the designated format consumable by the machine learning model.
3 . The computing system of claim 2 , wherein the sub-set of two or more prediction computing nodes is configured to output different real-time delay predictions for different current aircraft flights in parallel.
4 . The computing system of claim 2 , wherein one or more computing nodes of the plurality of computing nodes is a training computing node configured to train the machine learning model with the training data including the training flight-plan parameters, the training aircraft surveillance parameters, and the training weather parameters from the plurality of previously-completed aircraft flights.
5 . The computing system of claim 2 , wherein one or more computing nodes of the plurality of computing nodes is a cluster manager computing node configured to dynamically manage a designation of each of the plurality of computing nodes to operate as a data consumption computing node, a prediction computing node, or a training computing node based at least on a total number of current aircraft flights being monitored by the computing system.
6 . The computing system of claim 2 , wherein the plurality of different data feeds includes an airport data feed, wherein the data consumption computing node is configured to extract, from the airport data feed, an airport delay parameter for an airport associated with the respective current aircraft flight, wherein the aircraft information message includes the airport delay parameter, and wherein the machine learning model is configured to assess the real-time delay prediction for the respective current aircraft flight based at least on the airport delay parameter.
7 . The computing system of claim 1 , wherein the one or more aircraft flight-plan parameters include one or more of an aircraft type, an origin, a destination, a flight path, and an estimated time of arrival of the respective current aircraft flight.
8 . The computing system of claim 1 , wherein the one or more aircraft surveillance parameters include one or more of a current latitude, a current longitude, a current altitude, a current heading, and a current speed of the respective current aircraft flight.
9 . The computing system of claim 1 , wherein the one or more weather parameters include one or more of historical weather reports and a short-term weather forecast along a flight path of the respective current aircraft flight.
10 . A computer-implemented method comprising: receiving an aircraft information message for a respective current aircraft flight, the aircraft information message having a designated format consumable by a machine learning model previously trained to assess delay predictions for aircraft flights, the aircraft information message including one or more aircraft flight-plan parameters, one or more aircraft surveillance parameters, and one or more weather parameters for the respective current aircraft flight, wherein the machine learning model is previously trained with training data including training flight-plan parameters, training aircraft surveillance parameters, and training weather parameters from a plurality of previously-completed aircraft flights, wherein the training data further includes actual arrival times for the plurality of previously-completed aircraft flights and delay predictions for the plurality of previously-completed aircraft flights that include delay predictions for previously-completed aircraft flights that arrived ahead of schedule relative to an expected arrival time and delay predictions for previously-completed aircraft flights that arrived behind schedule relative to an expected arrival time, and wherein the machine learning model is previously trained using weighted training in which a delay prediction having a smaller error relative to a corresponding actual delay is weighted more than a delay prediction having a larger error relative to a corresponding actual delay; providing the aircraft information message as input to the machine learning model to assess a real-time delay prediction for the respective current aircraft flight based at least on the one or more aircraft flight-plan parameters, the one or more aircraft surveillance parameters, and the one or more weather parameters included in the aircraft information message wherein the real-time delay prediction for the respective current aircraft flight indicates a predicted arrival time of the aircraft and an indication whether the aircraft is ahead of schedule or behind schedule relative to an expected arrival time; outputting the real-time delay prediction for the respective current aircraft flight in a format that is consumable by a distributed event streaming platform that is configured to publish a data feed of real-time delay predictions for the respective current aircraft flight; receiving, via the distributed event streaming platform, a plurality of data feeds of real-time delay predictions for a plurality of current aircraft flights, including the real-time delay prediction for the respective current aircraft flight; extracting a plurality of real-time delay predictions for the plurality of current aircraft flights from the plurality of data feeds, including the real-time delay prediction for the respective current aircraft flight; and visually presenting the plurality of real-time delay predictions for the plurality of current aircraft flights in a graphical user interface (GUI).
11 . The computer-implemented method of claim 10 , wherein the aircraft information message is generated by extracting the one or more aircraft flight-plan parameters from an aircraft flight-plan data feed, the one or more aircraft surveillance parameters from an aircraft surveillance data feed, and the one or more weather parameters from a weather data feed for the respective current aircraft flight.
12 . The computer-implemented method of claim 11 , wherein the aircraft information message further includes an airport delay parameter extracted from an airport data feed for an airport associated with the respective current aircraft flight, and wherein the machine learning model is previously trained to assess the real-time delay prediction for the respective current aircraft flight based at least on the airport delay parameter.
13 . The computer-implemented method of claim 10 , wherein the one or more aircraft flight-plan parameters include one or more of an aircraft type, an origin, a destination, a flight path, and an estimated time of arrival of the respective current aircraft flight.
14 . The computer-implemented method of claim 10 , wherein the one or more aircraft surveillance parameters include one or more of a current latitude, a current longitude, a current altitude, a current heading, and a current speed of the respective current aircraft flight.
15 . The computer-implemented method of claim 10 , wherein the one or more weather parameters include one or more of historical weather reports and a short-term weather forecast along a flight path of the respective current aircraft flight.
16 . A computing system comprising: a plurality of computing nodes, each computing node of the plurality of computing nodes including a logic processor and a non-volatile storage device holding instructions executable by the logic processor; wherein a first set of one or more computing nodes of the plurality of computing nodes are designated as a data consumption computing node configured to: receive a plurality of different data feeds including an aircraft flight-plan data feed, an aircraft surveillance data feed, and a weather data feed, extract one or more aircraft flight-plan parameters, one or more aircraft surveillance parameters, and one or more weather parameters from the plurality of different data feeds for a respective current aircraft flight, and generate an aircraft information message for the respective current aircraft flight, wherein the aircraft information message has a designated format consumable by a machine learning model previously trained to assess delay predictions for aircraft flights, the aircraft information message including the one or more aircraft flight-plan parameters, the one or more aircraft surveillance parameters, and the one or more weather parameters for the respective current aircraft flight, wherein the machine learning model is previously trained with training data including training flight-plan parameters, training aircraft surveillance parameters, and training weather parameters from a plurality of previously-completed aircraft flights, wherein the training data further includes actual arrival times for the plurality of previously-completed aircraft flights and delay predictions for the plurality of previously-completed aircraft flights that include delay predictions for previously-completed aircraft flights that arrived ahead of schedule relative to an expected arrival time and delay predictions for previously-completed aircraft flights that arrived behind schedule relative to an expected arrival time, and wherein the machine learning model is previously trained using weighted training in which a delay prediction having a smaller error relative to a corresponding actual delay is weighted more than a delay prediction having a larger error relative to a corresponding actual delay; wherein a second set of one or more computing nodes of the plurality of computing nodes are designated as a prediction computing node configured to: receive the aircraft information message for the respective current aircraft flight, provide the aircraft information message as input to the machine learning model to assess a real-time delay prediction for the respective current aircraft flight based at least on the one or more aircraft flight-plan parameters, the one or more aircraft surveillance parameters, and the one or more weather parameters included in the aircraft information message, wherein the real-time delay prediction for the respective current aircraft flight indicates a predicted arrival time of the aircraft and an indication whether the aircraft is ahead of schedule or behind schedule relative to an expected arrival time, and output the real-time delay prediction for the respective current aircraft flight in a format that is consumable by a distributed event streaming platform that is configured to publish a data feed of real-time delay predictions for the respective current aircraft flight; wherein a third set of one or more computing nodes of the plurality of computing nodes are designated as a training computing node configured to: train the machine learning model with the training data; and wherein a fourth set of one or more computing nodes of the plurality of computing nodes is configured to: receive, via the distributed event streaming platform, a plurality of data feeds of real-time delay predictions for a plurality of current aircraft flights, including the real-time delay prediction for the respective current aircraft flight; extract a plurality of real-time delay predictions for the plurality of current aircraft flights from the plurality of data feeds including the real-time delay prediction for the respective current aircraft flight; and visually present the plurality of real-time delay predictions for the plurality of current aircraft flights in a graphical user interface (GUI).
17 . The computing system of claim 16 , wherein a fifth set of one or more computing nodes of the plurality of computing nodes is designated as a cluster manager computing node configured to dynamically manage a designation of each of the plurality of computing nodes to operate as a data consumption computing node, a prediction computing node, or a training computing node based at least on a total number of current aircraft flights being monitored by the computing system.
18 . The computing system of claim 1 , wherein the prediction computing node is configured to repeatedly receive aircraft information messages during the respective current aircraft flight, provide the aircraft information messages as input to the machine learning model to repeatedly assess real-time delay predictions for the respective current aircraft flight based at least on one or more aircraft flight-plan parameters extracted from the aircraft information messages, and output the real-time delay predictions for the respective current aircraft flight during the respective current aircraft flight to the distributed event streaming platform.
19 . The computer-implemented method of claim 10 , further comprising: repeatedly receiving aircraft information messages during the respective current aircraft flight; providing the aircraft information messages as input to the machine learning model to repeatedly assess real-time delay predictions for the respective current aircraft flight based at least on one or more aircraft flight-plan parameters extracted from the aircraft information messages; and outputting the real-time delay predictions for the respective current aircraft flight during the respective current aircraft flight to the distributed event streaming platform.
20 . The computing system of claim 16 , wherein the second set of one or more computing nodes of the plurality of computing nodes that are designated as prediction computing nodes are configured to repeatedly receive aircraft information messages during the respective current aircraft flight, provide the aircraft information messages as input to the machine learning model to repeatedly assess real-time delay predictions for the respective current aircraft flight based at least on one or more aircraft flight-plan parameters extracted from the aircraft information messages, and output the real-time delay predictions for the respective current aircraft flight during the respective current aircraft flight to the distributed event streaming platform.

Description

FIELD The present disclosure relates generally to the field of aircraft flight traffic monitoring, and more specifically to providing real-time flight-delay predictions of aircraft flights in a large-scale environment. BACKGROUND In recent years, the volume of aircraft flights has significantly increased causing corresponding increases in requirements for monitoring and management of such aircraft flights. However, existing aircraft flight monitoring and management systems have not been able to keep pace with the increased volume. One primary issue is a lack of accuracy in predicting when an aircraft flight will arrive at a destination. Such a lack of flight-delay prediction accuracy leads to inefficient management of airport resources. Further, such a lack of flight-delay prediction accuracy leads to increases in the number and extent of flight delays at airports, thereby negatively affecting wait times experienced by passengers. Further still, such a lack of flight-delay prediction accuracy has a negative economic impact on airline companies that have to compensate passengers for overbooking due to missed aircraft flights. SUMMARY To address the above and other issues, examples are disclosed that relate to providing real-time flight-delay predictions of airborne flights in a large-scale environment. In one example, an aircraft information message for a current aircraft flight is received. The aircraft information message has a designated format consumable by a machine learning model previously trained to assess delay predictions for aircraft flights. The aircraft information message includes one or more aircraft flight-plan parameters, one or more aircraft surveillance parameters, and one or more weather parameters for the current aircraft flight. The aircraft information message is provided as input to the machine learning model to assess a real-time delay prediction for the current aircraft flight based at least on the one or more flight-plan parameters, the one or more aircraft surveillance parameters, and the one or more weather parameters included in the aircraft information message. The features, functions, and advantages that have been discussed can be achieved independently in various embodiments or may be combined in yet other embodiments, further details of which can be seen with reference to the following description and drawings. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 shows an example scenario of an aircraft in a tactical phase of flight. FIG. 2 shows an example computing system configured to provide real-time flight-delay predictions of airborne flights in a large-scale environment. FIG. 3-4 show example computer architectures of a plurality of computing nodes for provide real-time flight-delay predictions of airborne flights in a large-scale environment. FIG. 5 shows an example aircraft information message having a designated format that is consumable by a machine learning model previously trained to assess delay predictions for aircraft flights. FIG. 6 shows an example graphical user interface configured to visually present real-time flight-delay predictions. FIG. 7-9 show example methods for providing real-time flight-delay predictions of airborne flights in a large-scale environment. DETAILED DESCRIPTION In the recent years there have been attempts to improve the accuracy of flight-delay predictions using different mathematical formulations of aircraft performance models. Existing aircraft performance models are based on equations of motion aggregated from historical flight data and are mainly ruled by initial and boundary conditions. A drawback to such an approach is the lack of a system definition for applying these existing aircraft performance models in a real-time data-driven environment, since the existing aircraft performance models are configured to process, filter, and analyze historical data and are not applicable in a real-time environment. FIG. 1 shows an example scenario of an aircraft 100 in a tactical phase of flight. In particular, the aircraft 100 has taken off from an origin airport 102 and is currently in the tactical phase of the flight in which the aircraft 100 is traveling to a destination airport. As discussed above, the accuracy of existing aircraft performance models for predicting flight delays is negatively affected by a lack of information regarding aspects that influence the development of the aircraft flight in this tactical phase. For example, existing aircraft performance models can lack information relating to weather disruptions, air-traffic sector congestions, air traffic controller actions, and any other disruptive event that can take place during the tactical phase of the flight. Such a lack of information considered by existing aircraft performance models results in flight-delay predictions that can have a significant error depending on the events that occur during the tactical phase of a flight. Accordingly, the present description is directed to a