Robotic agricultural remediation

{
  "publication_number": "US11553634B2",
  "country": "US",
  "kind": "B2",
  "title": "Robotic agricultural remediation",
  "abstract": "Implementations are described herein for analyzing vision data depicting undesirable plants such as weeds to detect various attribute(s). The detected attribute(s) of a particular undesirable plant may then be used to select, from a plurality of available candidate remediation techniques, the most suitable remediation technique to eradicate or otherwise eliminate the undesirable plants.",
  "claims": [
    "1. A method for robotic remediation of one or more plants, the method implemented by one or more processors and comprising: deploying one or more robots amongst a plurality of plants; acquiring, using one or more vision sensors of one or more of the robots, vision data that depicts one or more of the plurality of plants; processing the vision data using a machine learning model to classify first and second plants of the plurality of plants as targets for extermination; further analyzing the vision data to detect a first set of one or more attributes of the first plant and a second set of one or more attributes of the second plant; selecting a first remediation technique to exterminate the first plant, wherein the first remediation technique is selected from a plurality of candidate remediation techniques based on the first remediation technique being more suitable for exterminating plants having the one or more detected attributes of the first set than the other candidate remediation techniques, and wherein at least one the plurality of candidate remediation techniques is implementable autonomously by one or more of the robots; and selecting a second remediation technique to exterminate the second plant, wherein the second remediation technique is different than the first remediation technique and is selected from the plurality of candidate remediation techniques based on the second remediation technique being more suitable for exterminating plants having the one or more detected attributes of the second set than the other candidate remediation techniques.",
    "2. The method of claim 1, wherein the one or more attributes include a maturity level of the given plant.",
    "3. The method of claim 1, wherein the first set of one or more attributes includes proximity of the first plant to one or more adjacent plants.",
    "4. The method of claim 1, wherein the one or more attributes include a genus or species of the given plant.",
    "5. The method of claim 1, wherein at least two of the plurality of candidate remediation techniques are implementable autonomously by one or more of the robots.",
    "6. The method of claim 5, wherein the at least two of the plurality of candidate remediation techniques are selected from the following: incinerating the given plant with a single beam of coherent light; incinerating the given plant with multiple beams of coherent light; destroying the given plant with voltage transmitted through ionized air; and mechanically destroying the plant with a solid mass launched at a high velocity.",
    "7. The method of claim 6, wherein the solid mass is constructed with controlled-release fertilizer.",
    "8. The method of claim 6, wherein the solid mass comprises a gelatin-filled capsule.",
    "9. The method of claim 1, further comprising: calculating a likelihood that applying one or more available remediation techniques to the first plant will damage one or more nearby desirable plants; wherein selecting the first remediation technique is conditional based on the likelihood.",
    "10. The method of claim 1, further comprising: calculating a likelihood that applying one or more available remediation techniques to the first plant will damage one or more nearby desirable plants; wherein the first remediation technique is selected based on the likelihood.",
    "11. A robot comprising: one or more processors; and non-transitory computer-readable memory operably coupled with the one or more processors, the memory storing instructions that, in response to execution of the instructions by one or more of the processors, cause the one or more processors to: acquire, using one or more vision sensors of the robot, vision data that depicts one or more of a plurality of plants; process the vision data using a machine learning model to classify first and second plants of the plurality of plants as targets for extermination; further analyze the vision data to detect a first set of one or more attributes of the first plant and a second set of one or more attributes of the second plant; select a first remediation technique to exterminate the first plant, wherein the first remediation technique is selected from a plurality of candidate remediation techniques based on first remediation technique being more suitable for exterminating plants having the one or more detected attributes of the first set than the other candidate remediation techniques, and wherein at least one the plurality of candidate remediation techniques is implementable autonomously by the robot; and select a second remediation technique to exterminate the second plant, wherein the second remediation technique is different than the first remediation technique and is selected from the plurality of candidate remediation techniques based on the second remediation technique being more suitable for exterminating plants having the one or more detected attributes of the second set than the other candidate remediation techniques.",
    "12. The robot of claim 11, wherein the one or more attributes include a maturity level of the given plant.",
    "13. The robot of claim 11, wherein the first set of one or more attributes includes proximity of the first plant to one or more adjacent plants.",
    "14. The robot of claim 11, wherein the one or more attributes include a genus or species of the given plant.",
    "15. The robot of claim 11, wherein at least two of the plurality of candidate remediation techniques are implementable autonomously by the robot.",
    "16. The robot of claim 15, wherein the at least two of the plurality of candidate remediation techniques are selected from the following: incinerating the given plant with a single beam of coherent light; incinerating the given plant with multiple beams of coherent light; destroying the given plant with voltage transmitted through ionized air; and mechanically destroying the plant with a solid mass launched at a high velocity.",
    "17. The robot of claim 16, wherein the solid mass is constructed with controlled-release fertilizer.",
    "18. The robot of claim 16, wherein the solid mass comprises a gelatin-filled capsule.",
    "19. The robot of claim 11, further comprising instructions to calculate a likelihood that applying one or more available remediation techniques to the first plant will damage one or more nearby desirable plants; wherein the first remediation technique is selected conditionally based on the likelihood.",
    "20. At least one non-transitory computer-readable medium comprising instructions that, in response to execution of the instructions by one or more processors, cause the one or more processors to perform the following operations: deploying one or more robots amongst a plurality of plants; acquiring, using one or more vision sensors of one or more of the robots, vision data that depicts one or more of the plurality of plants; processing the vision data using a machine learning model to classify first and second plants of the plurality of plants as targets for extermination; further analyzing the vision data to detect a first set of one or more attributes of the first plant and a second set of one or more attributes of the second plant; selecting a first remediation technique to exterminate the first plant, wherein the first remediation technique is selected from a plurality of candidate remediation techniques based on the first remediation technique being more suitable for exterminating plants having the one or more detected attributes of the first set than the other candidate remediation techniques, and wherein at least one the plurality of candidate remediation techniques is implementable autonomously by one or more of the robots; and selecting a second remediation technique to exterminate the second plant, wherein the second remediation technique is different than the first remediation technique and is selected from the plurality of candidate remediation techniques based on the second remediation technique being more suitable for exterminating plants having the one or more detected attributes of the second set than the other candidate remediation techniques."
  ],
  "description_excerpt": "Weeds and other undesirable plants constitute a major cost for the agricultural industry. Many weeds can be treated with herbicides, but for a variety of reasons there is increasing resistance to their use. Herbicides can be expensive, especially when applied in a non-targeted, blanket manner. Herbicides are also coming under increasing environmental scrutiny for their role in contaminating groundwater and other water sources. There is a perception that chemicals used in agriculture may be harmful to human health from extended exposure, putting pressure on growers to stop using them. Additionally, with the increasing popularity of organic farming - for which non-organic herbicides are not typically permitted - organic herbicides may be cost-ineffective and/or inefficient. And many undesirable plants are developing resistance to herbicides.\n\nWith recent advances in artificial intelligence it has become practical to identify and classify individual plants as undesirable based on various types of sensor data, particularly two-dimensional (“2D”) vision data. For example, robots can travel through a field acquiring vision data that can be analyzed to identify/classify undesirable plants such as weeds or other interlopers. However, weeds come in numerous different types, morphologies, sizes, maturity levels, and/or arrangements relative to other, desired plants. This makes weed management challenging for robots, which are typically better at homogenous tasks.\n\nImplementations disclosed herein are directed to analyzing vision data depicting undesirable plants such as weeds to detect various attribute(s).",
  "cpc": [
    "A01B 39/18",
    "A01B 41/00",
    "A01B 79/02",
    "A01D 34/008",
    "A01M 21/043",
    "B25J 9/1679",
    "B25J 9/1697",
    "G05D 1/0246",
    "G05D 2201/0201",
    "G06T 2207/30188",
    "G06T 7/70",
    "G06V 20/64"
  ],
  "ipc": [
    "A01B 39/18",
    "A01D 34/00",
    "A01M 21/04",
    "G05D 1/02",
    "G06T 7/70",
    "G06V 20/64"
  ],
  "assignees": [
    "X Development LLC"
  ],
  "inventors": [
    "Elliott Grant",
    "Hongxiao Liu",
    "Zhiqiang YUAN",
    "Sergey Yaroshenko",
    "Benoit Schillings",
    "Matt VanCleave"
  ],
  "filing_date": "2019-10-01",
  "publication_date": "2023-01-17",
  "grant_date": "2023-01-17",
  "priority_date": "2019-10-01",
  "application_number": "US-201916589833-A",
  "family_id": "72944256",
  "cited_by_count": 9,
  "citations": [
    "US5442552A",
    "US7854108B2",
    "US9756771B2",
    "US9658201B2",
    "US10219449B2",
    "US20170034986A1",
    "US20160029612A1",
    "US9913429B1",
    "WO2017002093A1",
    "US20210112704A1",
    "EP3279831A1",
    "US20180330166A1",
    "US20210084813A1",
    "US20220044030A1"
  ]
}