Management device for cultivation of fruit vegetable plants and fruit trees, learning device, management method for cultivation of fruit vegetable plants and fruit trees, learning model generation method, management program for cultivation of fruit vegetable plants and fruit trees, and learning model generation program

Patent metadata

Field	Value
Publication number	US-12250910-B2
Application number	US-202017298729-A
Country	US
Kind code	B2
Filing date	Sep 29, 2020
Priority date	Oct 4, 2019
Publication date	Mar 18, 2025
Grant date	Mar 18, 2025

Abstract

A cultivation and management device for fruit vegetable plants and fruit trees includes environment state information on an environment state of a fruit vegetable plant or a fruit tree to be cultivated, planned cultivation evaluation index information on a preplanned cultivation evaluation index of the fruit vegetable plant or the fruit tree, a calculation unit configured to determine and output a work including a shape change work for the fruit vegetable plant or the fruit tree with respect to inputs of the environment state information and the planned cultivation evaluation index information with a learning model, and an output unit configured to output the work including the shape change work for the fruit vegetable plant or the fruit tree.

First claim

The invention claimed is:

1. A cultivation and management system for fruit vegetable plants and fruit trees comprising:
a greenhouse configured to house a fruit vegetable plant or a fruit tree to be cultivated;
at least one environment sensor configured to measure a value corresponding to an environment state of the greenhouse;
at least one growth sensor configured to measure a value corresponding to a growth state of the fruit vegetable plant or the fruit tree disposed in the greenhouse;
at least one environment state controller configured to control the environment state of the greenhouse;
a cultivation and management device including (i) at least one memory configured to store computer-executable instructions and at least one processor configured to execute the computer-executable instructions stored in the at least one memory, (ii) at least one integrated circuit, or both (i) and (ii) that cause the cultivation and management device to:
acquire environment state information indicating an environment state of the fruit vegetable plant or the fruit tree disposed in the greenhouse based on the value, measured by the at least one environment sensor, corresponding to the environment state of the greenhouse;
acquire growth state information indicating the growth state of the fruit vegetable plant or the fruit tree disposed in the greenhouse based on the value, measured by the at least one growth sensor, corresponding to the growth state of the fruit vegetable plant or the fruit tree;
acquire planned cultivation evaluation index information on a preplanned cultivation evaluation index of the fruit vegetable plant or the fruit tree;
implement a calculator configured to determine and output a work including a shape change work for the fruit vegetable plant or the fruit tree and being quantified according to a work type with respect to inputs of the environment state information, the growth state information, and the planned cultivation evaluation index information, the calculator using a learning model trained on a cultivation evaluation index of the fruit vegetable plant or the fruit tree cultivated, an environment state of the fruit vegetable plant or the fruit tree when cultivating the fruit vegetable plant or the fruit tree, a growth state of the fruit vegetable plant or the fruit tree when cultivating the fruit vegetable plant or the fruit tree, and a chronologically identified work history of a quantified historical work for the fruit vegetable plant or the fruit tree, the quantified historical work including the shape change work for changing a shape of the fruit vegetable plant or the fruit tree when cultivating the fruit vegetable plant or the fruit tree and being quantified according to the work type; and
implement an outputter configured to output the work including the shape change work for the fruit vegetable plant or the fruit tree and being quantified according to the work type,
wherein the calculator is further configured to use the learning model to determine and output a state control instruction with respect to the inputs of the environment state information, the growth state information, and the planned cultivation evaluation index information,
wherein the outputter is further configured to output the state control instruction to thereby control the at least one environment state controller to thereby control the environment state of the greenhouse, and
wherein the learning model is generated by reinforcement learning using the environment state information as a state of an environment, the quantified historical work including the shape change work for the fruit vegetable plant or the fruit tree as an action on the environment, and the cultivation evaluation index as a reward.

2. The cultivation and management system for fruit vegetable plants and fruit trees according to claim 1 , wherein
the chronologically identified work history includes a history of a quantified historical work whose amount of work cannot be expressed as a continuous value.

3. A cultivation and management method using a cultivation and management system for fruit vegetable plants and fruit trees, the cultivation and management system including a housing configured to house a fruit vegetable plant or a fruit tree to be cultivated, at least one environment sensor configured to measure a value corresponding to an environment state of the housing, at least one growth sensor configured to measure a value corresponding to a growth state of the fruit vegetable plant or the fruit tree disposed in the housing, and at least one environment state controller configured to control the environment state of the housing, the cultivation and management method comprising:
acquiring environment state information indicating an environment state of the fruit vegetable plant or the fruit tree disposed in the housing based on the value, measured by the at least one environment sensor, corresponding to the environment state of the housing;
acquiring growth state information indicating the growth state of the fruit vegetable plant or the fruit tree disposed in the housing based on the value, measured by the at least one growth sensor, corresponding to the growth state of the fruit vegetable plant or the fruit tree;
acquiring planned cultivation evaluation index information on a preplanned cultivation evaluation index of the fruit vegetable plant or the fruit tree;
determining a work including a shape change work for the fruit vegetable plant or the fruit tree and being quantified according to a work type with respect to inputs of the environment state information, the growth state information, and the planned cultivation evaluation index information, the work being determined using a learning model trained on a cultivation evaluation index of the fruit vegetable plant or the fruit tree cultivated, an environment state of the fruit vegetable plant or the fruit tree when cultivating the fruit vegetable plant or the fruit tree, a growth state of the fruit vegetable plant or the fruit tree when cultivating the fruit vegetable plant or the fruit tree, and a chronologically identified work history of a quantified historical work for the fruit vegetable plant or the fruit tree, the quantified historical work including the shape change work for changing a shape of the fruit vegetable plant or the fruit tree when cultivating the fruit vegetable plant or the fruit tree and being quantified according to the work type;
outputting the work including the shape change work for the fruit vegetable plant or the fruit tree and being quantified according to the work type;
determining, using the learning model, a state control instruction with respect to the inputs of the environment state information, the growth state information, and the planned cultivation evaluation index information; and
outputting the state control instruction to thereby control the at least one environment state controller to thereby control the environment state of the housing,
wherein the learning model is generated by reinforcement learning using the environment state information as a state of an environment, the quantified historical work including the shape change work for the fruit vegetable plant or the fruit tree as an action on the environment, and the cultivation evaluation index as a reward.

4. The cultivation and management method for fruit vegetable plants and fruit trees according to claim 3 , wherein
the chronologically identified work history includes a history of quantified historical work whose amount of work cannot be expressed as a continuous value.

5. A non-transitory computer readable medium storing a cultivation and management program for executing a cultivation and management method using a cultivation and management system for fruit vegetable plants and fruit trees

Assignees

• Omron Tateisi Electronics Co

Inventors

• Hosomi Shinichi
• Sasaki Sho
• Hashimoto Atsushi
• Miyaura Hiroyuki

Classifications

• G06N3/092

  Reinforcement learning · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/0499

  Feedforward networks · CPC title

• G06Q50/02

  Agriculture; Fishing; Forestry; Mining · CPC title

• G06Q10/06313

  Resource planning in a project environment · CPC title