Granted Patent B2

US 12,187,418 · App. 17/788,164 · Granted Jan 7, 2025

Inventor: George Howell (Bristol, GB)

Assignee: AIRBUS OPERATIONS LIMITED

B64C25/426B60T8/174G06N3/02

Abstract

Claims (60)

1. An aircraft control system comprising:

an input interface adapted to receive aircraft operating inputs and generate input data representing an operating state of an aircraft;

a processing engine comprising a classifier having a plurality of parameters adapted to represent a control policy for operating the aircraft, the processing engine arranged to apply the input data to the classifier to generate control output data, wherein the control policy represents a desired operation of the aircraft under a plurality of scenarios and the parameters are generated based on training data representing the control policy for each of the plurality of scenarios; and

an output interface adapted to generate control outputs from the control output data to control the aircraft,

wherein the training data is generated by selecting an operating procedure for the aircraft control system for each of the plurality of scenarios by:

determining a target outcome measure for the scenario;

generating a plurality of operating procedures for the scenario based on an environment representing at least part of the aircraft, wherein the environment includes:

a graph model representing the at least part of the aircraft, the graph model comprising:

a plurality of nodes including a start node and at least one end node, and

a set of directed edges connecting the plurality of nodes,

wherein the plurality of operating procedures are generated by determining pathways in the graph model from the start node to the at least one end node;

determining outcome measures based on the plurality of operating procedures; and

selecting an operating procedure from the plurality of operating procedures based on a comparison of the target outcome measure and with the outcome measures.

2. The aircraft control system according to claim 1 , wherein the aircraft operating inputs comprise signals from at least one of sensor outputs and user inputs associated with an operating state of the aircraft.

3. The aircraft control system according to claim 1 , wherein the environment representing the at least part of an aircraft is dependent on the said scenario.

4. The aircraft control system according to claim 1 , wherein the plurality of nodes comprises a first subset of the nodes, wherein each node in the first subset representing a piece of equipment being operated in the least part of an aircraft, and

a second subset of the nodes, wherein each node in the second subset representing a manner in which a piece of equipment is operated in the at least part of an aircraft.

5. The aircraft control system according to claim 1 , wherein the selecting an operating procedure based at least on a comparison of the target outcome measure with the outcome measures comprises:

selecting a subset of the operating procedures based on the comparison of the target outcome measure with the outcome measures;

generating additional operating procedures based on the subset of the operating procedures;

determining additional outcome measures from the additional operating procedures; and

selecting one of the additional an operating procedure based on a comparison of the target outcome measure with the additional outcome measures.

6. The aircraft control system according to claim 1 , wherein the aircraft control system is a braking control system.

7. The aircraft control system according to claim 4 , wherein the aircraft control system is a braking control system and the graph model represents actuation equipment and avionic equipment in the braking control system.

8. The aircraft control system according to claim 7 , wherein the nodes in the first subset of the nodes each represent operation of actuation equipment in the braking control system.

9. The aircraft control system according to claim 7 , wherein at least one node in the second subset of the nodes represent a manner in which a brake is operated.

10. The aircraft control system according to claim 9 , wherein the manner in which the brake is operated includes one or more of:

specifying avionic equipment used to operate said brake;

operating the brake in an environmental compensation mode; or

operating the brake not in an environmental compensation mode.

11. The aircraft control system according to claim 6 , wherein the input interface is configured to receive inputs from one or more of:

a power supply for operating the brake;

at least one of a plurality of coils used to operate respective components for actuation equipment in the braking control system;

a pressure sensor in the braking control system;

avionic equipment in the braking control system; or

a wheel speed sensor for monitoring braking.

12. A machine learning system for training a classifier of an aircraft control system, the machine learning system comprising:

an environment comprising a graph model representing at least part of an aircraft, the graph model having a plurality of nodes and a set of directed edges connecting the plurality of nodes, the plurality of nodes including a start node and at least one end node; and

a pathway evaluation engine adapted to:

obtain data representing at least one target outcome measure for a given scenario;

generate a plurality of pathways from the start node to at least one end node for the given scenario;

determine outcome measures based on the plurality of pathways;

select a pathway of the plurality of pathways based at least on a comparison of the target outcome measure with the outcome measures; and

generate training data representing a control policy for the given scenario based on the selected pathway;

storage for storing the training data to be used to train a classifier of an aircraft control system; and

a training engine adapted to train the classifier of the aircraft control system using the training data to generate control output data according to the control policy using input data representing an operating state of an aircraft.

13. The machine learning system according to claim 12 , wherein selecting the pathway comprises:

evaluating a further plurality of the pathways from the start node to at least one end node, wherein the further plurality of the pathways are generated based on a selected subset of the plurality of the pathways; and

selecting a pathway from the further plurality of pathways based on a comparison of at least one outcome measure determined for the selected pathway and the at least one target outcome measure.

14. The machine learning system according to claim 13 , wherein the selected subset of the plurality of the pathways are selected based on a comparison of outcome measures determined for the plurality of the pathways with the at least one target outcome measure.

15. An aircraft comprising the aircraft control system in claim 1 .

16. A method for training a classifier of an aircraft control system, the method comprising:

generating training data representing a control policy for a plurality of scenarios; and

training a classifier of the aircraft control system to generate control output data according to the control policy using input data representing an operating state of an aircraft;

wherein generating training data comprises, for each of the plurality of scenarios:

representing at least part of an aircraft using a graph model comprising a plurality of nodes and a set of directed edges connecting the plurality of nodes, the plurality of nodes comprising a start node and at least one end node;

obtaining data representing at least one target outcome measure for the scenario;

generating a plurality of pathways from the start node to the at least one end node for the plurality of scenarios;

determining outcome measures based on the plurality of pathways; and

selecting a pathway of the plurality of pathways based on a comparison of the at least one target outcome measure with the outcome measures.

Assignments (1)

Priority Claims (1)

Continuity (1)

References Cited (19)