Safe and secure practical autonomy

Patent metadata

Field	Value
Publication number	US-10656643-B1
Application number	US-201715629548-A
Country	US
Kind code	B1
Filing date	Jun 21, 2017
Priority date	Jun 21, 2017
Publication date	May 19, 2020
Grant date	May 19, 2020

Abstract

Safe practical autonomy is ensured by encapsulating an unreliable or untrusted machine learning algorithm within a control-based algorithm. A safety envelope is utilized to ensure that the machine learning algorithm does not output control signals that are beyond safe thresholds or limits. Secure practical autonomy is ensured by verification using digital certificates or cryptographic signatures. The verification may be for individual partitions of an autonomous system or apparatus. The partitions include trusted and untrusted partitions. Trusted partitions are verified for security, while untrusted partitions are verified for safety and security.

First claim

What is claimed:

1. An apparatus for safe and secure practical autonomy, comprising:
a controller having a non-transitory memory with instructions thereon and a processor in communication with the non-transitory memory, the processor configured to access the instructions;
at least one port for receiving a real-time operational parameter for generating a navigational solution including a platform maneuver based on the real-time operational parameter and a position;
a machine learning module communicatively coupled to the controller, wherein the instructions of the controller provide the processor access to a machine learning algorithm of the machine learning module to generate the navigational solution and direct the platform maneuver based on the position; and a safety monitor implementing a certifiable control algorithm and an operational safety envelope to preempt the machine learning module when the platform maneuver is outside the operational safety envelope, the operational safety envelope formed from a baseline or threshold associated with each control command of the machine learning module, the operational safety envelope is formed from a first real-time parameter used together in combination with a second real-time parameter, wherein the first real-time parameter is a radar return and the second real-time parameter is a separation distance threshold whereby the radar return is used with the separation distance threshold to establish the operational safety envelope.

2. The apparatus of claim 1 , wherein the real-time operational parameter includes an observable parameter provided to the machine learning module as an input, and wherein the machine learning module directs the platform based on the position and the observable parameter.

3. The apparatus of claim 1 , wherein the controller utilizes a cryptographic signature to ensure security of the machine learning algorithm and the certifiable control algorithm.

4. The apparatus of claim 1 , wherein the certifiable control algorithm is certifiable according to Federal Agency of Aviation (FAA) standards, European Aviation Safety Agency (EASA) standards, or industry standards such as the American National Standards Institute (ANSI).

5. The apparatus of claim 1 , wherein the operational safety envelope comprises a limit, a threshold, a baseline, or an appropriate safety level determined using two or more real-time parameters.

6. The apparatus of claim 1 , further comprising an autopilot sub-system for performing the platform maneuver, wherein the autopilot sub-system is in communication with the safety monitor and is implemented in a sub-system designed to a design assurance level (DAL) higher than a DAL associated with the machine learning module.

7. The apparatus of claim 6 , wherein the platform maneuver is a first maneuver and the autopilot sub-system is coupled to the safety monitor via a switch to perform a second maneuver when the machine learning module is preempted.

8. The apparatus of claim 1 , wherein the first real-time parameter used together in combination with the second real-time parameter includes a limit, threshold, baseline, or appropriate safety level at least two of maximum/minimum safe altitudes, allowable single-axis control parameters for maintaining flight with respect to a roll axis, allowable two-axis control parameters for maintaining flight with respect to a pitch axis and the roll axis, or allowable three-axis control parameters for maintaining flight with respect to the roll axis, the pitch axis, and a yaw axis.

Assignees

• Rockwell Collins Inc

Inventors

• Bertram Joshua R
• Wolford Brian R
• Mclean Angus L
• Roup Alexander V

Classifications

• G06N20/00

  Machine learning · CPC title

• B64C13/18

  using automatic pilot · CPC title

• G05D1/0088

  characterized by the autonomous decision making process, e.g. artificial intelligence, predefined behaviours (using knowledge based models G06N5/00) · CPC title

• G08G5/0013

  Physics · mapped topic

• G05D1/101

  specially adapted for aircraft · CPC title