Prognostics and health management system

Patent metadata

Field	Value
Publication number	US-10124893-B1
Application number	US-201715708013-A
Country	US
Kind code	B1
Filing date	Sep 18, 2017
Priority date	Sep 18, 2017
Publication date	Nov 13, 2018
Grant date	Nov 13, 2018

Abstract

Techniques are described for assessing the health of an unmanned vehicle such as an unmanned aerial vehicle (UAV). In some embodiments, sensors corresponding to subsystems of the UAV may be utilized to assess the health of a particular subsystem. Predictive models may be stored within memory of the UAV to enable such assessments to be performed at the UAV (e.g., during performance of a mission). Sensor data collected from sensors on the UAV may be provided as input for a predictive model associated with a particular subsystem. The predictive model may output a failure prediction indicating a likelihood, and in some cases, a time by which failure of the subsystem is predicted to occur given the sensor data. In some embodiments, one or more corrective actions may be identified and triggered based, at least in part, on the failure prediction.

First claim

What is claimed is:

1. An unmanned aerial vehicle (UAV), comprising:
one or more subsystems to at least generate lift and facilitate flight of the UAV;
one or more sensor devices associated with at least one of the one or more subsystems;
one or more memories configured to store a mission plan and a plurality of predictive models, the mission plan comprising at least one stage corresponding to a set of operations for executing the at least one stage of the mission plan, the plurality of predictive models individually being associated with a particular subsystem of the one or more subsystems; and
an onboard prognostics module communicatively coupled with the one or more subsystems and one or more sensor devices, wherein the onboard prognostics module is configured to:
identify a subsystem associated with a first stage of the mission plan;
monitor a set of sensor devices associated with the subsystem during performance of the first stage of the mission plan;
receive sensor data collected by the set of sensor devices associated with the subsystem, the sensor data being collected during performance of the first stage of the mission plan;
identify a predictive model of the plurality of predictive models based at least in part on the received sensor data, the predictive model being associated with the subsystem;
determine, based at least in part on the predictive model and sensor data, a time at which a failure of the subsystem is likely to occur;
determine a corrective action based at least in part on the time at which the failure of the subsystem is likely to occur, the corrective action being associated with modifying the mission plan; and
trigger the performance of the corrective action based at least in part on the time at which the failure of the subsystem is likely to occur, the corrective action being associated with modifying the mission plan.

2. The unmanned aerial vehicle (UAV) of claim 1 , wherein the first stage of the mission plan is related to performing flight operations using the one or more subsystems, and wherein a second stage of the mission plan is related to executing operations for delivering an item by the UAV.

3. The unmanned aerial vehicle (UAV) of claim 1 , wherein the onboard prognostics module is further configured to:
store collective sensor data within the one or more memories, the collective sensor data being collected by the one or more sensor devices during performance of the mission plan;
determine that the mission plan is concluded; and
transmit, to a ground management system, the collective sensor data.

4. The unmanned aerial vehicle (UAV) of claim 1 , wherein the onboard prognostics module is further configured to:
maintain the plurality of predictive models, wherein each of the predictive models corresponds to a specific subsystem of the one or more subsystems, wherein each of the plurality of predictive models accepts as input particular sensor data of a subset of the one or more sensor devices, and wherein each of the plurality of predictive models provides an output that relates to failure of the specific subsystem.

5. One or more computer-readable media comprising instructions that, when executed by an onboard prognostics module of an unmanned aerial vehicle (UAV), cause the onboard prognostics module to at least:
identify a subsystem of a the UAV, the subsystem being associated with a stage of a mission plan;
obtain sensor data collected by one or more sensors associated with the subsystem, the sensor data being collected during performance of the stage of the mission plan;
identify a predictive model based at least in part on the sensor data obtained, the predictive model being associated with the subsystem and stored within one or more memories of the UAV; and
cause one or more corrective actions to be performed by the UAV based at least in part on a determination that a failure of the subsystem is likely to occur, the determination being based at least in part on the predictive model and the sensor data.

6. The one or more computer-readable media of claim 5 , wherein the onboard prognostics module is further configured to:
obtain a plurality of predictive models, wherein each of the predictive models corresponds to a specific subsystem of one or more subsystems;
determine that the stage of the mission plan is currently being performed;
select the predictive model from the plurality of predictive models, the selection being based on the subsystem associated with the stage;
provide, as input to the selected predictive model, the sensor data collected by the one or more sensors associated with the subsystem during performance of the stage of the mission plan; and
receive output from the selected predictive model based at least in part on the input.

7. The one or more computer-readable media of claim 6 , wherein the output from the selected predictive model comprises at least a time at which the predictive model predicts the failure of the subsystem is to occur, wherein the one or more corrective actions are further cased based at least in part on the time at which the predictive model predicts the failure of the subsystem is to occur.

8. The one or more computer-readable media of claim 7 , wherein the output from the selected predictive model comprises at least a value indicating a likelihood that the failure occurs at the time at which the predictive model predicts the failure of the subsystem is to occur.

9. The one or more computer-readable media of claim 5 , wherein at least one of the subsystems facilitates item delivery by the UAV.

10. The one or more computer-readable media of claim 6 , wherein the plurality of predictive models are initially trained using historical sensor data collected from the UAV and a plurality of UAVs.

11. The one or more computer-readable media of claim 5 , wherein the onboard prognostics module is further configured to:
generate an updated predictive model based at least in part on the sensor data obtained, the updated predictive model being initially generated based at least in part on historical sensor data collected by a plurality of UAVs.

12. The one or more computer-readable media of claim 5 , wherein causing one or more corrective actions to be performed by the UAV further comprises:
obtaining a plurality of historical corrective actions previously performed by a plurality of UAVs;
identifying a set of potential corrective actions of the plurality of historical corrective actions based at least in part on the sensor data and output provided by the predictive model;
selecting the one or more corrective actions from the set of potential corrective actions; and
transmitting data related to the one or more corrective actions to a management module of the UAV, wherein receipt by the management module causes the one or more corrective actions to be performed by the UAV.

13. A computer-implemented method, comprising:
identifying, by a computing device of an unmanned aerial vehicle (UAV), a subsystem associated with a stage of a mission plan;
monitoring, by the computing device, one or more sensors associated with the subsystem during performance of the stage of the mission plan;
receiving, by the computing device, sensor data collected by the one or more sensor;
updating, by the computing device, a predictive model based at least in part on the received sensor data, the predictive model being associated with the subsystem and stored within a memory of the UAV;
determining, by the computing device, based at least in part on the predictive model and the sensor data, a time at which a failure of the subsystem is likely to occur;
determining, by the

Assignees

• Amazon Tech Inc

Inventors

• Aalund Ryan Jeffrey
• Alonso Lopez Javier

Classifications

• B64U2101/60

  for transporting passengers; for transporting goods other than weapons · CPC title

• B64U2201/20

  Remote controls · CPC title

• B64U2201/10

  autonomous, i.e. by navigating independently from ground or air stations, e.g. by using inertial navigation systems [INS] · CPC title

• B64C39/024 Primary

  of the remote controlled vehicle type, i.e. RPV · CPC title

• G07C5/008

  communicating information to a remotely located station (transmission systems for measured values G08C) · CPC title