Granted Patent B2

US 12,664,462 · App. 17/571,015 · Granted Jun 23, 2026

Inventors: Michael Krenz (Roscoe, IL); Darin Michael Underwood (Cedar Rapids, IA)

Assignee: Hamilton Sundstrand Corporation

G06N20/00

Abstract

Claims (37)

1 . A real-time artificial intelligence and/or machine learning (AI/ML) system for an aircraft, comprising:

an AI/ML module configured to receive one or more inputs and to calculate an AI/ML control output, wherein the AI/ML module includes a non-deterministic model for processing the one or more inputs and outputting the AI/ML control output; and

a deterministic module configured to receive the one or more inputs and the AI/ML control output from the AI/ML module, wherein the deterministic module includes a deterministic model for processing the one or more inputs and/or the AI/ML control output to calculate a deterministic condition, wherein the deterministic module is configured to check the AI/ML control output against the deterministic condition to determine whether the AI/ML control output satisfies the deterministic condition;

wherein the deterministic module comprises a reverse computation checker such that the deterministic module is configured to calculate reverse-computed inputs based on the AI/ML control output, wherein the deterministic module is configured to compare the reverse-computed inputs to the one or more inputs, and wherein the deterministic module is configured to output the AI/ML control output only if the reverse-computed inputs are equal to or within a set range of the one or more inputs.

2 . The system of claim 1 , wherein the AI/ML module is unqualified according to one or more regulations with respect to aircraft, and wherein the deterministic module is qualified according to the one or more regulations.

3 . The system of claim 1 , wherein the deterministic module further comprises a boundary condition checker such that the deterministic module is configured to calculate one or more boundary conditions for the AI/ML control output based on the one or more inputs, and wherein the deterministic module is configured to output the AI/ML control output only if the AI/ML control output satisfies the one or more boundary conditions.

4 . The system of claim 3 , wherein the deterministic module is configured to output a modified output that is a boundary closest to the AI/ML control output when the AI/ML control output is outside of the one or more boundary conditions.

5 . The system of claim 4 , wherein the deterministic module is configured to output the modified output that is a set output or deterministic output when the reverse-computed inputs are not equal to or outside the set range of the one or more inputs.

6 . The system of claim 1 , wherein the deterministic module further comprises a command limiter such that the deterministic module is configured to calculate one or more command limits for the AI/ML control output based on the one or more inputs, and wherein the deterministic module is configured to output the AI/ML control output only if the AI/ML control output is at or within the one or more command limits.

7 . The system of claim 6 , wherein the one or more command limits are rate limits.

8 . The system of claim 6 , wherein the deterministic module is configured to output a modified output that is a command limit closest to the AI/ML control output when the AI/ML control output is not at or within the one or more command limits.

9 . An aircraft, comprising:

a real-time artificial intelligence and/or machine learning (AI/ML) system, comprising:

an AI/ML module configured to receive one or more inputs and to calculate an AI/ML control output, wherein the AI/ML module includes a non-deterministic model for processing the one or more inputs and outputting the AI/ML control output; and

a deterministic module configured to receive the one or more inputs and the AI/ML control output from the AI/ML module, wherein the deterministic module includes a deterministic model for processing the one or more inputs and/or the AI/ML control output to calculate a deterministic condition, wherein the deterministic module is configured to check the AI/ML control output against the deterministic condition to determine whether the AI/ML control output satisfies the deterministic condition;

wherein the deterministic module comprises a reverse computation checker such that the deterministic module is configured to calculate reverse-computed inputs based on the AI/ML control output, wherein the deterministic module is configured to compare the reverse-computed inputs to the one or more inputs, and wherein the deterministic module is configured to output the AI/ML control output only if the reverse-computed inputs are equal to or within a set range of the one or more inputs.

10 . The aircraft of claim 9 , wherein the AI/ML module is unqualified according to one or more regulations with respect to aircraft, and wherein the deterministic module is qualified according to the one or more regulations.

11 . The aircraft of claim 9 , wherein the deterministic module further comprises a boundary condition checker such that the deterministic module is configured to calculate one or more boundary conditions for the AI/ML control output based on the one or more inputs, and wherein the deterministic module is configured to output the AI/ML control output only if the AI/ML control output satisfies the one or more boundary conditions.

12 . The aircraft of claim 11 , wherein the deterministic module is configured to output a modified output that is a boundary closest to the AI/ML control output when the AI/ML control output is outside of the one or more boundary conditions.

13 . The aircraft of claim 12 , wherein the deterministic module is configured to output the modified output that is a set output or deterministic output when the reverse-computed inputs are not equal to or outside the set range of the one or more inputs.

14 . The aircraft of claim 9 , wherein the deterministic module further comprises a command limiter such that the deterministic module is configured to calculate one or more command limits for the AI/ML control output based on the one or more inputs, and wherein the deterministic module is configured to output the AI/ML control output only if the AI/ML control output is at or within the one or more command limits.

15 . The aircraft of claim 14 , wherein the one or more command limits are rate limits.

16 . The aircraft of claim 14 , wherein the deterministic module is configured to output a modified output that is a command limit closest to the AI/ML control output when the AI/ML control output is not at or within the one or more command limits.

17 . A non-transitory computer readable medium having computer executable instructions configured to cause a computer to perform a method, the method comprising:

receiving, at an artificial intelligence and/or machine learning (AI/ML) module of an aircraft, one or more inputs;

calculating and outputting an AI/ML control output using AI/ML based on the one or more inputs;

receiving, at a deterministic module, the one or more inputs and the AI/ML control output from the AI/ML module;

processing the one or more inputs and/or the AI/ML control output to calculate a deterministic condition;

checking the AI/ML control output against the deterministic condition to determine whether the AI/ML control output satisfies the deterministic condition;

calculating reverse-computed inputs based on the AI/ML control output using a reverse computation checker implemented by the deterministic module;

comparing the reverse-computed inputs to the one or more inputs using the deterministic module; and

outputting the AI/ML control output only if the reverse-computed inputs are equal to or within a set range of the one or more inputs using the deterministic module.

18 . The non-transitory computer readable medium of claim 17 , wherein:

using the AI/ML includes using a non-deterministic model to process the one or more inputs and output the AI/ML control output, the non-deterministic model unqualified according to one or more regulations with respect to aircraft; and

calculating the deterministic condition includes using a deterministic model, the deterministic model qualified according to the one or more regulations.

19 . The system of claim 1 , wherein the deterministic module is further configured to output the AI/ML control output in response to determining that the AI/ML control output satisfies the deterministic condition and to output a modified output in response to determining that the AI/ML control output does not satisfy the deterministic condition.

20 . The aircraft of claim 9 , wherein the deterministic module is further configured to output the AI/ML control output in response to determining that the AI/ML control output satisfies the deterministic condition and to output a modified output in response to determining that the AI/ML control output does not satisfy the deterministic condition.

Assignments (1)

Continuity (1)

References Cited (12)