Unmanned aerial vehicle with neural network for enhanced mission performance
An unmanned aerial vehicle (UAV) or “drone” executes a neural network to assist with inspection, surveillance, reporting, and other missions. The drone inspection neural network may monitor, in real time, the data stream from a plurality of onboard sensors during navigation to an asset along a preprogrammed flight path and/or during its mission (e.g., as it scans and inspects an asset).
1. An unmanned aerial vehicle (UAV) comprising:
a flight package;
a navigation system;
an image-acquisition device;
a communication facility;
a computer memory; and
a computer including a processor and electronically stored instructions, executable by the processor, for:
causing the UAV to autonomously execute a preliminary flight plan around an asset to be inspected,
using data received from the image-acquisition device as input to a neural-network-based predictor that has been computationally trained to identify and classify objects appearing in images acquired by the image-acquisition device during the preliminary flight plan,
during analysis of the data received from the image-acquisition device, detecting a feature requiring an update to the predictor,
requesting and receiving, in real time, an updated predictor computationally trained based on the detected feature, and
based on (a) object classifications made by the predictor during the preliminary flight plan and (b) the updated predictor, computing and autonomously executing a revised flight plan around the asset
wherein:
the computer is configured to analyze the detected feature in real-time and classify the detected feature as normal or anomalous; and
when the detected feature is classified as anomalous, the computer is configured to request airspace authorization and receive updated configuration files for additional inspection of the asset.
2. The UAV of claim 1 , wherein the communication facility is configured to interact with terrestrial and airborne control systems.
3. The UAV of claim 1 , further comprising a database of actions, the computer being configured to select and cause execution of an action from the database in response to a detected object classified by the predictor.
4. The UAV of claim 1 , further comprising a weather-surveillance module for monitoring weather conditions during flight of the UAV.
5. The UAV of claim 1 , wherein the computer is further configured to communicate with a high-altitude pseudosatellite (HAPS) vehicle and to execute flight commands received from the HAPS vehicle.
6. The UAV of claim 1 , wherein the detected feature is an asset type, and the updated predictor comprises weights files, configuration files, and/or convolutional layers specific to the asset type.
7. The UAV of claim 1 , wherein the detected feature is passive intermodulation (PIM) or electromagnetic interference (EMI).
8. The UAV of claim 1 , wherein the computer is configured to conduct real-time quality assurance of data acquired during execution of the revised flight plan and, based thereon, determine whether to continue the revised flight plan or acquire additional data.
9. The UAV of claim 1 , wherein the detected feature is classified as an anomaly and wherein the computer is configured to request and receive the updated predictor in response to detection of the feature requiring the update to the predictor.
10. The UAV of claim 5 , wherein the computer is further configured to communicate the revised flight plan to the HAPS vehicle for obtaining authorization from air-traffic control infrastructure.
11. The UAV of claim 5 , wherein the computer is further configured to communicate the request for the updated predictor to the HAPS vehicle and receive the updated predictor from the HAPS vehicle.
12. The UAV of claim 5 , wherein the computer is configured to dynamically adjust computing power of the predictor via communication with the HAPS vehicle and/or a cloud neural network module.
13. A method of inspecting an asset using an unmanned aerial vehicle (UAV), the method comprising the steps of:
acquiring digital images in real time during an autonomous flight of the UAV in accordance with a preliminary flight plan around an asset to be inspected;
computationally analyzing the acquired digital images with a neural-network-based predictor that has been computationally trained to identify and classify objects appearing in the images;
during the analysis of the acquired digital images, computationally detecting a feature requiring an update to the predictor;
while the UAV is in flight, requesting and receiving, in real time, an updated predictor computationally trained based on the detected feature; and
while the UAV is in flight, and based on (a) object classifications made by the predictor during the preliminary flight plan and (b) the updated predictor, computing and autonomously executing a revised flight plan around the asset,
the method further comprising:
analyzing the detected feature in real-time and classifying the detected feature as normal or anomalous; and
when the detected feature is classified as anomalous, requesting airspace authorization and receiving updated configuration files for additional inspection of the asset.
14. The method of claim 13 , wherein revised flight plan is computed based at least in part on database lookup in response to a detected object classified by the predictor.
15. The method of claim 13 , further comprising monitoring weather conditions while the UAV is in flight, the revised flight plan being further based on the monitored weather conditions.
16. The method of claim 13 , further comprising the step of communicating the revised flight plan to a high-altitude pseudosatellite (HAPS) vehicle for obtaining authorization from air-traffic control infrastructure.
17. The method of claim 13 , wherein the detected feature is an asset type, and the updated predictor comprises weights files, configuration files, and/or convolutional layers specific to the asset type.
18. The method of claim 13 , further comprising conducting real-time quality assurance of data acquired during execution of the revised flight plan and, based thereon, determining whether to continue the revised flight plan or acquire additional data.
19. The method of claim 13 , wherein the detected feature is passive intermodulation (PIM) or electromagnetic interference (EMI).
20. The method of claim 13 , wherein the detected feature is classified as an anomaly and wherein the step of requesting and receiving the updated predictor is performed in response to detection of the feature requiring the update to the predictor.
21. The method of claim 16 , further comprising dynamically adjusting computing power of the predictor via communication with the HAPS vehicle and/or a cloud neural network module.