SYSTEM AND METHOD FOR CONTROLLING AN IRRIGATION SYSTEM ON A FARM

Abstract

The invention relates to a method and a control system (3) for an irrigation installation (1) on a farm (13), comprising a control computer (33) configured to interact with measuring devices (5) and actuators (9) of said irrigation installation (1) through a refined and aligned conversational AI, referred to as current conversational AI (35), to control the irrigation of said farm (13) based on data transmitted by the measuring devices (5) and taking into account information relating to the farm and the irrigation installation. Figure for the abstract: Fig. 1

Description

Title of the invention: SYSTEM AND METHOD FOR CONTROLLING AN IRRIGATION SYSTEM ON A FARM Technical field

[0001] The present invention relates to the field of irrigation of a farm and more particularly, an automatic irrigation taking into account the properties of the soils and the specificities of the crops. PREVIOUS STATE OF THE ART

[0002] The irrigation of a farm depends on many parameters, not all of which are measurable. One of the most important parameters is the measurement of soil moisture, which varies with depth.

[0003] Currently, there are known methods which allow the irrigation of a farm to be triggered from soil moisture thresholds.

[0004] One such method is tensiometry, which measures a pressure drop corresponding to the suction force that the root must exert to extract the available water. This method is useful for determining the amount of water available to the plant but requires several probes per reading level and regular maintenance to obtain reliable information.

[0005] The most widespread technique for measuring soil moisture is capacitive measurement. This is most often performed by measuring the frequency of an LC oscillator immersed in the soil. It should be noted that capacitance depends on permittivity and is therefore not a direct measure of moisture. Permittivity can be related to moisture through empirical formulas that depend on the medium and the operating frequency.

[0006] However, the capacitive measurement technique suffers primarily from the difficulty of calibration due to the multitude of influencing parameters. These influencing parameters include the diversity of soil textures; the variety of liquid / air / solid interfaces depending on the environment; the conductivity of the pore fluid (dissolved ions); temperature; compaction and homogeneity of the medium; cracks; and the presence and activity of organic matter (e.g., rhizosphere, mycelium, algae, earthworms). The estimation of these factors is limited by the knowledge of certain non-measurable parameters, such as, for example, the saturated hydraulic conductivity of the soil and the hydraulic roughness of the plot surface. Calibration also depends on the measurement frequency and the soil texture: a device operating at a high frequency will be less sensitive to conductivity but will have a smaller analysis volume.

[0007] Thus, the calibration of capacitive sensors proves to be complex, tedious, and specific to each type of soil. It is possible to reduce the influence of certain factors depending on the frequency, but various relaxation phenomena cover the entire frequency spectrum.

[0008] Furthermore, for a given moisture level, the water available to plants varies depending on the soil type. Most commercially available capacitive sensors operate at frequencies in the range of 10 MHz to 100 MHz. These high frequencies are sensitive to water absorbed by the soil particles, which plants cannot extract, and are less suitable for measuring the capillary water available for irrigation.

[0009] In addition, the irrigation conditions are characterized by the dose and duration of irrigation as a function of the available flow rate as well as other parameters relating, for example, to the specificity of the soil and its storage capacity which have impacts on plant growth according to their adaptation to the environment.

[0010] This multitude of influencing parameters only allows us to have an approximate view of the water profile compared with the recommendations themselves, which are generic.

[0011] Other irrigation techniques use empirical models based on meteorological data to control irrigation systems. These models are essentially based on evapotranspiration, which depends on temperature, atmospheric pressure, light intensity, wind, relative humidity, water quality, solar radiation, and crop type.

[0012] These empirical models remain very limited and can only give a static and very partial representation of irrigation conditions and the state of the crop.

[0013] Thus, the object of the present invention is to remedy the aforementioned drawbacks by proposing a method and an automatic control system for an irrigation installation that is simple to implement, taking into account the specificity of the soil and type of crop while being compatible with all measurement techniques and regardless of the measurement frequencies used. Description of the invention

[0014] The invention relates to a control system for an irrigation installation on a farm, comprising a control computer configured to interact with measuring devices and actuators of said irrigation installation through a refined and aligned conversational AI, referred to as current conversational AI, to control the irrigation of said farm based on the data transmitted by the measuring devices and taking into account information relating to agricultural operation and irrigation installation.

[0015] This enables automatic and intelligent control by optimizing results (water consumption, crop volume and quality) according to the specific characteristics of the farm. Furthermore, this AI-based system operates regardless of the sensor measurement frequency, allowing the use of low-power, low-cost sensors that do not require complex calibration. The AI ​​learns on its own to correct and compensate for sensor errors.

[0016] Advantageously, the pilot computer is configured to determine a current water profile based on changes in plant root depth.

[0017] This allows the amount of watering to be adjusted according to the maturity of the crop, which evolves according to the depth of the roots.

[0018] Advantageously, the flight computer is configured to perform preliminary learning on an initial conversational AI, said flight computer being configured to:

[0019] - to refine the initial conversational AI by providing it with scientific documents relating to agricultural irrigation, producing a refined conversational AI with theoretical knowledge about irrigation, and

[0020] - align the refined conversational AI by providing it with examples of its actual operation producing said current conversational AI having real knowledge about irrigation.

[0021] Thus, current conversational AI is a conversational AI refined and aligned for farm irrigation.

[0022] Advantageously, the control computer comprises:

[0023] -a conditioning module configured to condition the current conversational AI by providing it with said information relating to the farm operation and the irrigation installation, and

[0024] -an interface module configured for:

[0025] - to receive, at each clock-controlled sampling instant, measurement data from the measuring devices,

[0026] -transform said measurement data received from the measurement devices into textual data, and / or tokens, and / or lexical embeddings,

[0027] -return said textual data and / or tokens, and / or lexical embeddings to the current conversational AI which is configured to analyze and make a decision to start, continue or stop irrigation, and

[0028] -control at least one actuator of the irrigation system according to the decision made by the current conversational AI.

[0029] Thus, current conversational AI is evolving towards a model specifically made for the irrigation of the farm in question.

[0030] Advantageously, said information relating to the agricultural operation and the irrigation installation includes the type of soil, the crop, and the area of ​​the agricultural operation, as well as the types of measuring devices, the types of actuators, and the watering rate of the irrigation installation.

[0031] Advantageously, said measurement data include measurements of humidity on several soil levels, measurements of air humidity, measurements of temperature, brightness, wind, root depth, an estimate of the probability of precipitation, and a timestamp.

[0032] Advantageously, the control system includes a fine-tuning database. The interface module is configured to time-stampe and record in the fine-tuning database the history of its interactions with the measuring devices and actuators.

[0033] Thus, the entire discussion between the AI ​​and the sensors and actuators is logged. The sensors provide the measurements, the AI ​​provides the reasoning and the decisions (for example, hour by hour).

[0034] Advantageously, the piloting computer includes an annotation module configured to annotate the history of interactions, thus forming an annotated log in the fine-tuning database.

[0035] The annotation makes it possible to detect bad decisions and change the content of the record in order to make a better decision than the one that has already been made.

[0036] Advantageously, the pilot computer is configured to receive observations on the state of the plants and soil of the farm and in that the annotation module is configured to annotate the history recorded in the fine-tuning database according to said state of the plants according to their maturity, and of the soil.

[0037] Thus, the condition of the plants and the soil is a relevant indicator for AI which makes it unnecessary to know precisely the soil moisture level.

[0038] Advantageously, the piloting calculator is configured to evaluate a posteriori the effectiveness of irrigation piloting by annotating a set of quality indicators including the yield and quality of a harvest at the end of it.

[0039] This allows irrigation to be optimized while meeting user satisfaction criteria.

[0040] Advantageously, the control computer is configured to perform calibration of measurements from measuring devices by continuous learning using observations on the state of plants and soil on the farm.

[0041] This eliminates the need for complex and time-consuming calibration of capacitive sensors where the capacitive measurement is not a direct measurement of humidity.

[0042] Advantageously, the pilot computer is configured to continuously refine current conversational AI by reinjecting the current annotated log into a learning process of a LoRA-type LLM language model forming a current conversational AI model with LoRA adapter.

[0043] Thus, thanks to this technique, the current AI with LoRA adapter continues to learn from its logged and annotated experiences. Initially, the AI ​​was trained theoretically and on a few practical examples, and now, thanks to LoRA, the AI ​​possesses in-depth and practical knowledge of the specific field of agricultural operation.

[0044] The invention also relates to an irrigation system for a farm comprising the control system according to the preceding characteristics, further comprising:

[0045] -a set of measuring devices configured to take measurements on the plants, soil, and air of the farm, and to transmit the measurements to the control system, and

[0046] -a set of actuators configured to start or stop irrigation according to the command received from the control system.

[0047] Advantageously, the measuring device comprises a tube made of insulating material equipped with capacitive humidity measurement sensors, a power supply module, an acquisition and processing module and a communication module.

[0048] Advantageously, the irrigation installation includes an observation system comprising cameras configured to transmit images and / or observations on the condition of the plants and soil of the farm to the control system.

[0049] The invention also relates to a method of controlling an irrigation system on a farm, comprising a step of interaction with measuring devices and actuators of said irrigation system through a refined and aligned conversational AI, called current conversational AI, to control the irrigation of said farm according to the data transmitted by the measuring devices and taking into account information relating to the farm and the irrigation system.

[0050] Advantageously, the method includes a first learning phase comprising the following steps:

[0051] - to refine the initial conversational AI by providing it with scientific documents relating to agricultural irrigation, producing a refined conversational AI with theoretical knowledge about irrigation, and

[0052] -align refined conversational AI by providing it with examples of its actual operation producing said current conversational AI having real knowledge about irrigation.

[0053] Advantageously, the process includes a learning phase comprising the following steps:

[0054] -conditioning of current conversational AI by providing it with said information relating to agricultural operation and irrigation installation,

[0055] - to receive, at each clock-controlled sampling instant, measurement data from the measuring devices,

[0056] -transformation of said measurement data received from the measurement devices into textual data, and / or tokens, and / or lexical embeddings,

[0057] -returns said textual data, and / or tokens, and / or lexical embeddings, to the current conversational AI which is configured to analyze and make a decision to start, continue or stop irrigation, and

[0058] - control of at least one actuator of the irrigation system according to the decision made by the current conversational AI.

[0059] Other advantages and features of the invention will become apparent in the detailed, non-limiting description below. Brief description of the drawings

[0060] By way of non-limiting examples, embodiments of the invention will now be described, with reference to the accompanying drawings, in which:

[0061] Fig. 1 illustrates very schematically an irrigation installation on a farm, according to one embodiment of the invention;

[0062] Figure 2 schematically illustrates a measuring device intended for use in agricultural operations, according to one embodiment of the invention; and

[0063] Figs. 3A, 3B, 3C, 3D and 3E illustrate very schematically in relation to Fig. 1, the different phases of interaction between the pilot computer and the conversational artificial intelligence of the piloting system, according to a preferred embodiment of the invention. DETAILED DESCRIPTION OF THE INVENTION

[0064] The principle of the invention is to use conversational AI to automatically control the irrigation of a farm by continuously taking into account soil properties, crop specificity and observations on plants.

[0065] Fig. 1 illustrates very schematically an irrigation installation on a farm, according to one embodiment of the invention.

[0066] The irrigation installation 1 includes a control system 3, a set of measuring devices 5 for measuring soil moisture 7, a set of actuators 9 for starting or stopping the irrigation and optionally, an observation system 11.

[0067] The measuring devices 5 are intended to be partially embedded in the soil 7 of the farm in order to measure soil moisture as a function of depth. The measuring devices 5 are also configured to take measurements on the plants 12, the soil 7, and the air of the farm 13 and to transmit all measurements to the control system 3.

[0068] The observation system 11 includes cameras 14 and / or observation sensors to take images and / or observations on the condition of the plants 12 and the soil 7 of the farm 13. The observation system 11 may also include AI adapted to identify water stress in plants or signs of saturation in the images taken by the cameras 14. For example, water stress can be identified by wilting plants and signs of water saturation can be identified by puddles in the fields.

[0069] The cameras 14 and / or observation sensors can be installed in different locations of the farm 13. Alternatively, drones including cameras and / or observation sensors can also be used to monitor the development of the farm.

[0070] The observation system 11 based on cameras 14 and an adapted AI is thus adapted to transmit to the control system 3 images and / or observations on the state of the agricultural operation 13.

[0071] It should be noted that a primary source of observation is advantageously the farmer, who can provide the control system 3 with information on the condition of his farm. In this case, the observation system 11 can supplement the direct observations made by the farmer.

[0072] Fig. 2 illustrates very schematically a measuring device intended for use in agricultural operations, according to one embodiment of the invention.

[0073] According to this embodiment, the measuring device 5 comprises a tube 15 made of insulating material equipped with humidity measuring sensors 17, a power supply module 19, an acquisition and processing module 21 and a communication module 23.

[0074] The insulating material of the tube 15 can be PVC, PET, polycarbonate, glass, ceramic, etc. The length of the tube 15 depends on the depth of the plant's root system. The number of sensors 17 depends on the length of the tube 15 and therefore also on the depth of the plant's roots.

[0075] Each humidity measurement sensor 17 is advantageously a capacitive sensor comprising a resonant electrical circuit whose resonance frequency is representative of a permittivity measurement which is transformed via empirical formulas into a soil moisture measurement.

[0076] The resonant electrical circuit is an inductive-capacitive circuit comprising an active reactance associated with a corresponding passive coupling element. The active reactance is, for example, a pair of metallic armatures, and the passive coupling element is an inductive element. The geometry and configuration of the metallic armatures can be optimized according to the measurement range and the desired accuracy. An example of a capacitive sensor is described in patent application FR3115111.

[0077] Advantageously, the tube 15 made of insulating material is equipped with different types of capacitive sensors 17 having electrical circuits resonating at different frequencies, suitable for different types of crops and different sensitivities.

[0078] Optionally, the measuring device 5 may include or be associated with other sensors or probes including a temperature sensor 25, a conductivity measuring sensor 27, a weather station 29, a time clock 31, etc.

[0079] The temperature and conductivity sensors are advantageously distributed along the probe at each humidity measurement level in order to fuse these measurements because they are interdependent.

[0080] Advantageously, the power supply module 19 includes an autonomous means of energy harvesting such as, for example, a photovoltaic panel or a micro turbine. Alternatively, it may include simple electric batteries.

[0081] The acquisition and processing module 21 is, for example, a low-power microcontroller with a standby mode. The microcontroller 21 is configured to extract measurements and data from the various sensors 17, 25, 27, 29, 31 in order to determine soil moisture measurements as a function of depth, moisture and temperature gradient measurements as a function of depth, weather data, etc.

[0082] The communication module 23 is configured to communicate the results from the acquisition and processing module 21 with or without wires to the control system 3.

[0083] According to the invention, the control system 3 comprises a control computer 33 configured to interact with the measuring devices 5 and actuators 9 through a refined and aligned conversational artificial intelligence “AI”, referred to as current conversational AI 35. The control computer 33 is configured to automatically control the irrigation of the farm 13 based on the data transmitted by the measuring devices 5 and taking into account information relating to the farm 13 and the irrigation installation 1.

[0084] Irrigation control by the control computer 33 involves determining the optimal current water profile based on the evolution of plant root depth. Indeed, root depth changes with maturity, implying a change in the optimal water profile. It should be noted that data relating to the evolution of plant root depth are specific to each crop and can be found in the literature or obtained practically through sampling.

[0085] The control system 3 further includes databases 37 containing various data relating to general information on agricultural irrigation, information on the specific agricultural operation 13, real examples of the operation of the irrigation installation 1, fine-tuning data relating to the history of interactions of the current conversational AI 35 with the measuring devices 5 and actuators 9, etc.

[0086] Thanks to continuous learning by the current conversational AI 35, which uses information and observations relating to the farm 13 and the irrigation system 1, the control computer 33 is configured to calibrate the measurements from the measuring devices 5. This allows the system to operate at a frequency best suited to the specific characteristics of the crop. This eliminates the need for complex and costly calibration of the capacitive sensors 17, where the capacitive measurement is linked to a multitude of influencing parameters that cannot be directly measured.

[0087] Furthermore, the conversational AI-based processing works regardless of the measurement frequency. For example, measurements can be acquired at low frequencies, thereby increasing sensitivity to the water available to plants while simultaneously increasing the volume of measurements taken by the sensor.

[0088] Furthermore, thanks to the continuous learning of the current conversational AI, the measuring device can be equipped with the first 17a and second 17b types of capacitive sensors. The first capacitive sensors 17a have high-frequency resonant electrical circuits, for example, above approximately 30 MHz, in order to reduce the impact of the conductivity and texture of the soils in field crops. The second capacitive sensors 17b have low-frequency resonant electrical circuits, for example, between 0.1 MHz and 10 MHz, to be more sensitive to the conductivity and texture of the soils, which are factors in the availability of nutrients and water in the soil. Thus, regardless of the measurement frequency, the current conversational AI 35 allows the measurements from the measuring devices 5 to be calibrated so that they are indicative of soil moisture levels.

[0089] It should be noted that the control system 3 can be located locally on the farm. Alternatively, a farmer managing the farm can connect to a remote control system at a service provider or to a cloud.

[0090] Figs. 3A-3E illustrate very schematically in relation to [Fig.1], the different phases of interactions between the control computer 33 and the conversational artificial intelligence 35 of the control system 3, according to a preferred embodiment of the invention.

[0091] Fig. 3A represents a first phase in which the control computer 33 is configured to perform preliminary learning on an initial conversational AI 35a. Advantageously, the initial conversational AI 35a is a pre-trained conversational AI.

[0092] In step PI 1, the control computer 33 is configured to fine-tune the initial conversational AI 35a by providing it with scientific documents relating to agricultural irrigation from a first database 37a, thus producing a fine-tuned conversational AI 35b. Fine-tuning allows the conversational AI model to acquire preliminary theoretical knowledge about irrigation.

[0093] In step P12, the control computer 33 is configured to align the refined conversational AI 35b by providing it from a second database 37b with examples of its real operation including predetermined real interactions with the measuring devices 5 and actuators 9 of the irrigation installation 1. This alignment produces a refined and aligned conversational AI, called the current conversational AI 35. The alignment makes it possible to force the AI ​​model to behave according to concrete examples of operation.

[0094] This first phase thus allows the current conversational AI 35 to begin irrigation control by applying initial knowledge. Then, during subsequent phases, the current conversational AI 35 will learn progressively the consequences of its own tasks.

[0095] Fig. 3B is a second phase in which the control computer 33 is configured to evolve the current conversational AI 35 into a model specifically made for the irrigation of the farm 13 in question.

[0096] Indeed, the control computer 33 includes a conditioning module 41 and an interface module 43 configured to perform steps P21-P25. The interface module 43 is, for example, an application designed to communicate with the current conversational AI 35 and with the measuring devices 5 and actuators 9.

[0097] At step P21, the conditioning module 41 is configured to condition the current conversational AI 35 by providing it with information 44 relating to the farm and the irrigation system. The information relating to the farm 13 includes the soil type (clay, sand, etc.), the Crop (corn, sunflower, etc.), crop maturity, and farm area. Information relating to irrigation system 1 includes the types of sensors integrated into the measuring devices 5, the types of actuators 9, and the irrigation system's watering rate according to, for example, crop maturity, etc. Thus, the conditioning is a set of data enabling the current conversational AI 35 to have information about the ongoing task.

[0098] In addition, during steps P22-P25, the interface module 43 enables interaction through current conversational AI 35 with the measuring devices 5 and actuators 9. It also enables the timestamping and recording of these interactions.

[0099] More specifically, in step P22, the interface module 43 is configured to receive, at each sampling instant timed by a clock 45, measurement data from the measuring devices 5. This measurement data includes humidity measurements on several soil levels, air humidity measurements, temperature measurements, brightness, wind, root depth, an estimate of the probability of precipitation (for example within 24 hours), and a timestamp (i.e. the time of day and the period of the year).

[0100] At step P23, the interface module 43 is configured to transform the measurement data received from the measurement devices 5 into textual data and / or tokens and / or word embedding.

[0101] Thus, according to a first particular embodiment, the measurement data are transformed into textual data comprising character strings representing the measurement data and providing information on the current state of the measurement devices 5.

[0102] According to a first variant, the measurement data are transformed into tokens containing symbols corresponding to identifiers directly interpretable by current conversational AI.

[0103] According to a second variant, the measurement data are transformed into a lexical embedding by vectorizing the data into a latent vector space of the model. This abstract representation technique reduces the dimensionality of the word representation, thus facilitating learning. Lexical embeddings from several measurement devices can be hybridized by training an adaptation layer. This allows the current conversational AI model to be endowed with a 'sense' enabling it to directly 'understand' what it 'sees', what it 'hears', and what it 'feels' (temperature, humidity, etc.).

[0104] At step P24, interface module 43 is configured to return textual data, and / or tokens and / or lexical embeddings to the current conversational AI 35. The latter is configured to analyze and make a decision to start, continue or stop irrigation.

[0105] At step P25, the interface module 43 is configured to control at least one actuator 9 of the irrigation installation according to the decision taken by the current conversational AI at step P24.

[0106] It should be noted that safeguards can be implemented for this second phase in order to reduce the risk of bad decisions.

[0107] The [Fig.3C] is a third phase intended to prepare a refinement database.

[0108] Indeed, the control computer 33 also includes an annotation module 47 in addition to the conditioning modules 41 and interface module 43. During this third phase, the control computer 33 uses the interface module 43 and the annotation module 47 to prepare the fine-tuning database.

[0109] At step P31, the interface module 43 is configured to timestamp and log or record the history 46 of its interactions with the measuring devices 5 and actuators 9 through the current conversational AI 35 in a database 37c. The history 46 includes the reasoning and decisions taken by the current conversational AI 35, for example, hour by hour.

[0110] At step P32, the annotation module 47 is configured to annotate the history 46 of interactions recorded in the database 37c, thus forming a refinement database 37d, also called the 'annotated log'. More specifically, the condition of the plants and / or soil on the farm can be annotated by a farmer 53 or operator of that farm. The farmer 53 can optionally re-annotate what has been previously annotated.

[0111] Furthermore, the annotation module 47 is configured to use information 51 on the state of plants and / or soil from images and / or observations of the observation system 11 to annotate the history 46 of interactions recorded in the database 37c.

[0112] The annotation aims to detect poor decision-making and change the content of the recording or simply indicate that the result of a decision was not good in order to allow the current conversational AI 35 to progress by making better decisions than those previously made.

[0113] Advantageously, the annotation also allows the amount of watering to be adjusted according to the maturity of the crop.

[0114] Conversational AI current 35 can thus learn by itself by observing the differences between objectives and reality. After a few experiments, conversational AI current 35 discovers the correct decision-making. This can then be permanently added to the ongoing refinement of conversational AI.

[0115] As it operates, the conversational AI observes the effects of its decisions on the plants on the farm. For example, if the conversational AI observes that the plants are experiencing water stress, it may conclude that its previous reasoning was flawed. In this way, the conversational AI can learn from the consequences of its mistakes, enabling it to learn and improve its reasoning and future decisions.

[0116] Thus, the condition of the plants and the soil is used here as an indicator of water stress or over-irrigation either through observation by the farmer 53, or by means of cameras 14. This makes it possible to dispense with knowing precisely the soil moisture level knowing that this latter parameter is not a relevant indicator on its own, because for a type of soil with a given hydrometry profile, the capacity of plants to absorb moisture is not the same.

[0117] All these observations and annotations on the state of the plants and the soil allow current conversational AI 35 to further improve the calibration of measurements from measuring devices 5.

[0118] Advantageously, the control calculator 33 is configured to evaluate the effectiveness of irrigation control after the fact by annotating a set of quality indicators. The set of quality indicators includes the yield (e.g., water consumption, crop weight, growing time, diseases, etc.) and the quality (e.g., taste) of a crop at the end of the growing season.

[0119] In this case, the control system is configured to take these annotations into account by adjusting the initial parameters during a subsequent culture to meet user satisfaction criteria.

[0120] It should be noted that the user can assign a weight to each quality indicator according to the importance they place on each of these criteria. For example, growth rate may be less important than quality depending on the farmer's priorities. As harvests progress, the farmer may, for instance, seek to reduce water consumption while improving harvest volume and quality. In this case, high weights are assigned to these three criteria to identify the most favorable water profiles at each stage of crop development.

[0121] The water profile can be modified by the duration and frequency of watering, knowing that if we water for a long time and less often the moisture will be greater at depth than if we water little and often.

[0122] Adapting the water profile to the plant's needs throughout its growth optimizes water consumption while ensuring the quantity and quality of the harvest. Thus, feedback from past harvests allows us to converge towards the optimal water profile for each phase of growth.

[0123] The [Fig.3D] is a fourth phase intended to use the Large Language Model (LLM) technique to achieve further continuous refinement.

[0124] At step P41, the control computer 33 is configured to continuously refine the current conversational AI 35 by reinjecting the current annotated log 37d into a learning process of a LoRA (Low Rank Adaptation) type LLM language model. It should be noted that LoRA is a rapid, low-resource refinement technique that allows the entire model network to be blocked except for a few areas. That is, some layers are left free, allowing the current conversational AI not to forget what it has previously learned.

[0125] This step P41 is a loop in which the annotated log in the database 37d, which contains only correct or corrected decisions, is fed back by the control computer 33 into a LoRA-type learning process to refine the current conversational AI 35. This refinement of the current conversational AI 35 consists of taking the "previous version of the LoRA model" 61 and training on the accumulated data ("LoRA fine tuning") 63 to produce a new "current conversational AI model" with a LoRA component, referred to as "current conversational AI with LoRA adapter" 135. This same loop is repeated until the current conversational AI converges. It should be noted that at each iteration, certain layers of the current conversational AI with LoRA adapter 135 will be modified to take into account new information or experience.The LoRa adapter can be advantageously merged with the current conversational AI model.

[0126] Continuous fine-tuning thus enables lighter learning by using the log annotated from the beginning to update the AI ​​model. Using the LoRA technique, this type of fine-tuning allows updating only a few portions or LoRA layers of the current conversational AI model with the LoRA 135 adapter and is therefore much less demanding in terms of computing power required.

[0127] With this technique, current conversational AI with LoRA 135 adapter continues to learn from its experiences and should converge towards a highly specialized AI model of the agricultural operation in question.

[0128] The frequency of continuous fine-tuning may depend on poor decisions. As long as the current conversational AI with LoRA 135 adapter makes the right decisions, further fine-tuning is not necessary.

[0129] By way of example, this fourth phase can be carried out on a server of a supplier. Alternatively, it can also be carried out locally on the farm.

[0130] The [Fig.3E] is a fifth phase intended to use current conversational AI with LoRA adapter to better manage the irrigation of agricultural holdings.

[0131] This fifth phase is identical to the second phase of [Fig.3B], except that it is the current conversational AI with LoRA 135 adapter that is used for the farmer's operation.

[0132] Thus, at step P51, the conditioning module 41 is configured to condition the current conversational AI with LoRA adapter 135 by providing it with information relating to the farm operation and the irrigation installation.

[0133] At step P52, the interface module 43 is configured to receive, at each sampling instant timed by the clock 45, measurement data from the measuring devices 5.

[0134] At step P53, the interface module 43 is configured to transform the measurement data received from the measurement devices 5 into textual data and / or tokens and / or lexical embeddings.

[0135] At step P54, interface module 43 is configured to return textual data, and / or tokens and / or lexical embeddings to current conversational AI with LoRA adapter 135. The latter is configured to analyze and make a decision to start, continue or stop irrigation.

[0136] At step P55, the interface module 43 is configured to control at least one actuator 9 of the irrigation installation according to the decision taken by the current conversational AI with LoRA adapter 135 at step P54.

[0137] Thus, we continue to accumulate records that could eventually be used to train similar AIs for other farms. This allows the AI ​​to learn to adapt to climate change and to better manage the farm.

[0138] Advantageously, the control method or system can be applied to several consecutive crops to recover the language models of successive learnings in order to evaluate a posteriori the effectiveness of irrigation control.

[0139] This method or control system can also be applied to several crops in parallel to retrieve language models from the different training sessions. This makes it possible to create a generic conversational AI that can easily adapt to a new farm.

[0140] The control system according to the invention is very simple to implement while being compatible with all types of crops, soils, measurement methods or irrigation. It takes into account the farmer's knowledge, the specific characteristics of the soil, and the type of crop. It eliminates the need for complex calibration of capacitive sensors by considering indicators from both the plants and the soil. Furthermore, the control system is compatible with low-power probes and sensors.

[0141] Of course, various modifications can be made by a person skilled in the art to the invention just described, only by way of non-limiting examples.

Claims

Demands

1. Control system for an irrigation installation on a farm, characterized in that it comprises a control computer (33) configured to: -interact with measuring devices (5) and actuators (9) of said irrigation installation (1) through a refined and aligned conversational AI, called current conversational AI (35), made from preliminary learning on an initial conversational AI, and to control the irrigation of said farm (13) according to the data transmitted by the measuring devices (5) and taking into account information relating to the farm and the irrigation installation.

2. Control system according to claim 1, characterized in that the control computer (33) is configured to determine a current water profile as a function of a change in the root depth of plants.

3. A control system according to claim 1 or 2, characterized in that to carry out preliminary learning on the initial conversational AI, said control computer being configured to: -fine-tune the initial conversational AI by providing it with scientific documents relating to agricultural irrigation producing a refined conversational AI having theoretical knowledge about irrigation, and -align the refined conversational AI by providing it with examples of its actual operation producing said current conversational AI (35) having actual knowledge about irrigation.

4. A control system according to any one of claims 1 to 3, characterized in that the control computer (33) comprises: - a conditioning module (41) configured to condition the current conversational AI (35) by providing it with said information relating to the farm operation (13) and the irrigation installation, and - an interface module (43) configured to: -receive at each sampling instant timed by a clock (45), measurement data from the measurement devices (5), -transform said measurement data received from the measurement devices (5) into textual data, and / or tokens, and / or lexical embeddings, -return said textual data, and / or said tokens, and / or said lexical embeddings to the current conversational AI (35) which is configured to analyze and make a decision to start, continue or stop the irrigation, and -control at least one actuator (9) of the irrigation installation according to the decision made by the current conversational AI.

5. Control system according to claim 4, characterized in that said information relating to the farm (13) and the irrigation installation includes the type of soil, the crop, and the area of ​​the farm, as well as the types of measuring devices, the types of actuators, and the watering rate of the irrigation installation.

6. Control system according to claim 4, characterized in that said measurement data include humidity measurements on several soil levels, air humidity measurements, temperature measurements, brightness measurements, wind measurements, root depth measurements, an estimate of the probability of precipitation, and a timestamp.

7. A control system according to any one of claims 4 to 6, characterized in that it comprises a fine-tuning database, and in that the interface module (35) is configured to time-stamp and record in the fine-tuning database the history of its interactions with the measuring devices and actuators.

8. Control system according to claim 7, characterized in that it comprises an annotation module (47) configured to annotate the history of interactions thus forming an annotated log in the fine-tuning database.

9. A control system according to claim 8, characterized in that the control computer (33) is configured to receive observations on the condition of the plants and soil of the farm and in that the annotation module (47) is configured to annotate the history recorded in the refinement database according to the said state of the plants according to their maturity, and the soil.

10. Control system according to claim 8 or 9, characterized in that the control computer (33) is configured to evaluate a posteriori the effectiveness of irrigation control by annotating a set of quality indicators including the yield and quality of a harvest at the end thereof.

11. Control system according to claim 9 or 10, characterized in that the control computer (33) is configured to perform calibration of measurements from measuring devices (5) by continuous learning using observations on the state of plants and soil of the farm.

12. A control system according to any one of claims 8 to 11, characterized in that the control computer (33) is configured to continuously refine current conversational AI by reinjecting the current annotated log into a learning process of a LoRA-type LLM language model forming a current conversational AI model with LoRA adapter (135).

13. An irrigation installation for a farm comprising the control system according to any one of the preceding claims, characterized in that it further comprises: - a set of measuring devices (5) configured to take measurements on the plants, soil, and air of the farm (13), and to transmit the measurements to the control system (3), and - a set of actuators (9) configured to start or stop the irrigation according to the command received from the control system.

14. Irrigation installation according to claim 13, characterized in that a measuring device (5) comprises a tube (15) made of insulating material equipped with capacitive moisture measurement sensors (17a, 17b), a power supply module (19), an acquisition and processing module (21) and a communication module (23).

15. Irrigation installation according to claim 13 or 14, characterized in that it comprises an observation system (11) including cameras (14) configured to transmit to the control system (3) images and / or observations on the condition of the plants and soil of the farm.

16. Method for controlling an irrigation system on a farm, characterized in that it comprises the following steps: - interaction with measuring devices (5) and actuators (9) of said irrigation system (1) through a refined and aligned conversational AI, called current conversational AI (35), made from preliminary learning on an initial conversational AI, and - controlling the irrigation of said farm according to the data transmitted by the measuring devices and taking into account information relating to the farm and the irrigation system.

17. A piloting method according to claim 16, characterized in that it comprises a first learning phase comprising the following steps: - fine-tuning an initial conversational AI by providing it with scientific documents relating to agricultural irrigation producing a fine-tuned conversational AI having theoretical knowledge about irrigation, and - aligning the fine-tuned conversational AI by providing it with examples of its actual operation producing said current conversational AI having actual knowledge about irrigation.

18. A control method according to claim 16, characterized in that it comprises a learning phase including the following steps: - conditioning the current conversational AI by providing it with said information relating to the farm and the irrigation installation, - receiving at each clock-timed sampling instant, measurement data from the measuring devices, - transforming said measurement data received from the measuring devices into textual data, and / or tokens, and / or lexical embeddings, - sending said textual data, and / or tokens, and / or lexical embeddings back to the current conversational AI which is configured to analyze and make a decision to start, continue or stop the irrigation, and - controlling at least one actuator of the irrigation installation according to the decision made by the current conversational AI.

Images