# US 2019/0356392 A1 · Coherent aperture combining using machine learning > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 2019/0356392 A1 - Document ID: US20190356392A1 - Status: Published application - Kind: A1 - Application number: 16/387,015 - Filed: 2019-04-17 - Published: 2019-11-21 - Earliest claimed: 2017-12-28 - Assignees: Facebook, Inc. - Applicants: Facebook, Inc. - Inventors: Konstantin Kudinov; Amy Zhang; Tobias Gerard Tiecke; Kevin Michael Birnbaum - Web page: https://patentgrep.com/patent/US20190356392A1 ## CPC - H04B 10/1123: {Bidirectional transmission} - H04B 10/671: {for controlling the input optical signal} - G06T 7/74: {involving reference images or patches} - H04B 10/112: Line-of-sight transmission over an extended range - H04B 10/116: Visible light communication - H04B 10/697: {Arrangements for reducing noise and distortion} - G06T 2207/10016: Video; Image sequence - G06T 2207/20081: Training; Learning - G06T 2207/30168: Image quality inspection - H04B: TRANSMISSION - G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL ## Abstract An optical system for responding to distortions in incident light in a free space optical communication system includes a machine learning output storing at least an indication of multiple images and corresponding positioning or orientation attributes for one or more optical elements; a sensor configured to generate an image; and a component configured to adjust the one or more optical elements based on the generated image. Various other methods, systems, and apparatuses are also disclosed. ## Claims ### Claim 1 An optical system, comprising: a machine learning output that stores: an indication of multiple stored images; and at least one of positioning or orientation attribute values for one or more optical elements corresponding to the multiple stored images; a receiver aperture configured to receive a portion of an incident beam of light following propagation of the incident beam of light through a distorting medium; a sensor configured to generate an image of a footprint of the portion of the incident beam of light received by the receiver aperture; and a controller configured to adjust the one or more optical elements based on the generated image and a set of attribute values corresponding to at least one of the multiple stored images. ### Claim 2 (depends on 1) The optical system of claim 1, wherein the controller is further configured to adjust at least one of a position or an orientation of the one or more optical elements based on the generated image and the set of attribute values. ### Claim 3 (depends on 1) The optical system of claim 1, wherein: the controller is further configured to identify a closest stored image indicated by the multiple stored images of the machine learning output; and the set of attribute values corresponds to the closest stored image. ### Claim 4 (depends on 1) The optical system of claim 1, wherein: the controller is further configured to identify at least two closest stored images indicated by the multiple stored images of the machine learning output; and the set of attribute values is obtained by interpolating between attribute values corresponding to the at least two closest stored images. ### Claim 5 (depends on 1) The optical system of claim 1, wherein the generated image and at least some of the multiple stored images correspond to free space optical communications. ### Claim 6 (depends on 1) The optical system of claim 1, wherein a controller controls one or more actuators associated with the one or more optical elements. ### Claim 7 (depends on 1) The optical system of claim 1, wherein the one or more optical elements include at least one of a mirror, a beam splitter, or a lens. ### Claim 8 (depends on 1) The optical system of claim 1, further comprising a machine learning training module that generates the machine learning output based on a corpus of training images. ### Claim 9 (depends on 8) The optical system of claim 8, wherein the corpus of training images includes images having various amounts of distortion. ### Claim 10 (depends on 1) The optical system of claim 1, wherein the incident beam of light comprises a beam of coherent light transmitted from a separate location. ### Claim 11 (depends on 1) The optical system of claim 1, wherein the footprint of the portion of the incident beam of light received by the receiver aperture is less than the total footprint of the beam of light incident at the receiver aperture. ### Claim 12 (depends on 1) The optical system of claim 1, wherein the sensor comprises a plurality of light-sensing pixels. ### Claim 13 (depends on 12) The optical system of claim 12, further comprising mapping a displacement of an optical beam hotspot on the sensor to a displacement of the incident beam of light. ### Claim 14 A method, comprising: receiving a portion of an incident beam of light by a receiver aperture following propagation of the incident beam of light through a distorting medium; generating, by a sensor, an image of a footprint of the portion of the incident beam of light received by the receiver aperture; comparing the generated image to multiple stored images in a machine learning output to find at least one closest stored image; retrieving at least one of positioning or orientation attribute values for one or more optical elements indicated by the machine learning output; and adjusting, by a controller, the one or more optical elements based on the generated image and a set of attribute values corresponding to the at least one closest stored image. ### Claim 15 (depends on 14) The method of claim 14, wherein adjusting the one or more optical elements further comprises adjusting at least one of a position or an orientation of the one or more optical elements based on the generated image and the set of attribute values. ### Claim 16 (depends on 14) The method of claim 14, wherein: comparing the generated image to the multiple stored images further comprises comparing the generated image to the multiple stored images to find at least two closest stored images; and the set of attribute values is obtained by interpolating between attribute values corresponding to the at least two closest stored images. ### Claim 17 (depends on 14) The method of claim 14, further comprising receiving training images and generating the machine learning output by using a training function of one or more machine learning techniques. ### Claim 18 (depends on 17) The method of claim 17, wherein the training function includes supervised learning or unsupervised learning. ### Claim 19 (depends on 14) The method of claim 14, further comprising mapping a displacement of an optical beam hotspot on the sensor to a displacement of the incident beam of light. ### Claim 20 A computer-readable storage device storing computer-executable instructions, the computer-executable instructions comprising: instructions for receiving a portion of an incident beam of light by a receiver aperture following propagation of the incident beam of light through a distorting medium; instructions for generating, by a sensor, an image of a footprint of the portion of the incident beam of light received by the receiver aperture; instructions for comparing the generated image to multiple stored images in a machine learning output to find at least one closest stored image; instructions for retrieving at least one of positioning or orientation attribute values for one or more optical elements indicated by the machine learning output; and instructions for adjusting, by a controller, the one or more optical elements based on the generated image and a set of attribute values corresponding to the at least one closest stored image.