BE1030358A1 - METHOD FOR CONTROL OF AN UNMANNED AUTONOMOUS VEHICLE BASED ON A SOIL TYPE - Google Patents

Abstract

The present invention relates to a method for controlling an unmanned autonomous vehicle based on a soil type on a terrain, comprising determining the position of an unmanned autonomous vehicle, capturing an image of the terrain with a camera of the vehicle, image processing, creating a control signal, wherein the vehicle comprises at least one global and one local neural network, wherein the global network is trained with a global training set, comprising at least 70% images of other terrain, and the local network with a local training set, comprising images of the terrain, wherein one or more networks are selected from a group formed by said global and local networks based on the position, wherein during image processing each of the selected networks determines at least one classification of at least one soil type in the image, and wherein the control signal is basis of the classifications is created. The invention also relates to an unmanned autonomous vehicle and a use.

Description

1 BE2022/5191

METHOD FOR CONTROL OF AN UNMANNED AUTONOMOUS VEHICLE

GEAR BASED ON SOIL TYPE

TECHNICAL DOMAIN

The invention relates to a method for controlling an unmanned autonomous vehicle based on a soil type on a terrain, to an unmanned autonomous vehicle suitable for carrying out the method and to a use of the method and/or the unmanned autonomous vehicle for the autonomous maintenance of a garden.

STATE OF THE TECHNOLOGY

Such a method is known, among others, from WO 2018/220528 (WO '528).

WO '528 describes a method for controlling a soil tillage implement based on image processing. The soil working means comprises a drive means and a tool. In a first step of the method, at least one image of the soil is recorded using a digital acquisition device placed on the soil tillage equipment. Then, using a neural network, at least one convolution operation is performed on the image, resulting in at least one description of the bottom, after which a control signal for the drive or tool is obtained on the basis of the at least one description. . The neural network is trained using a multitude of images.

This known method has the disadvantage that it only carries out a certain action based on a determination of a soil type. This makes it possible that an action is performed at a specific location that is not desired there. For example, at a certain location the tillage implement could be allowed to drive over fallen leaves and perform a certain action, but at another location the fallen leaves could hide a soil type where this action is unnecessary or could even damage the tillage implement. . Another disadvantage of the method is that the neural network must be trained using an enormous number of images. Capturing all these images takes a lot of time, which means that the tillage product only needs to be used after a long period of recording

2 BE2022/5191 images can be used on a site, unless the neural network of the soil cultivation equipment has been trained in advance on the basis of an already existing collection of images. This has the additional disadvantage that the neural network is therefore not always optimally trained for describing soil types on the site where the tillage equipment is active. This can lead to incorrect description of the soil, whereupon incorrect actions can be carried out by the soil tillage equipment, again resulting in possible damage to the soil tillage equipment.

The aim of the invention is to provide a method that eliminates these disadvantages.

SUMMARY OF THE INVENTION

In a first aspect, the present invention concerns a method according to claim 1.

The advantage of this method is that the unmanned vehicle comprises both at least one global neural network and at least one local neural network. The at least one global neural network is advantageous because it makes the unmanned vehicle immediately deployable, without a user having to first create a training set of images of the entire terrain and train a neural network. This would require a lot of time from the user to capture a large number of images of the terrain and to annotate all these images. This is also expensive for a user because a lot of memory storage is required to store the training set. The at least one local neural network is advantageous to use the at least one local neural network to achieve a better classification in specific cases where determining at least one classification of at least one soil type in an image by the global network results in an incorrect classification. obtain. The at least one local neural network is additionally advantageous because only a lower number of images in the local training set is required to obtain a better classification in those specific cases, compared to retraining the at least one global neural network by adding images from the terrain to the global training set. This last solution would require a huge number of images to have any impact on the training of at least one global neural network for those specific cases, as the global training set already includes a plethora of images from a large number of different areas. Particularly advantageous is that one or more are based on the position of the unmanned autonomous vehicle

3 BE2022/5191 neural networks are selected from the group formed by the at least one global neural network and the at least one local neural network. This means that a user does not have to train a local neural network for the entire terrain, but only for those parts where a global neural network produces an incorrect classification, making it difficult to capture the images for a local training set and train a local neural network. network only requires a limited amount of time and limited memory. Because multiple global neural networks and/or multiple local neural networks can be selected, it is possible to make an optimal selection based on the desired classifications in a part of the terrain.

Preferred forms of the device are shown in claims 2 to 15.

A specific preferred form concerns a method according to claim 2.

An advantage of this embodiment is that the control signals are created on the basis of predetermined rules and that there are at least two sets of predetermined rules, whereby a set of predetermined rules is selected depending on the position of the unmanned autonomous vehicle. . This is advantageous because it allows the same selection of neural networks to be used in different parts of the terrain, so that the same classifications will be obtained for the same soil types, but a different control signal is still created. In this way, it can be avoided that an action that is not desired is carried out in a certain part of a site.

In a second aspect, the present invention concerns an unmanned autonomous vehicle according to claim 16.

Such an unmanned autonomous vehicle is advantageous because it can be immediately deployed on a site to perform tasks without additional training of a neural network by a user of the unmanned autonomous vehicle, while a local neural network can be used with limited effort by the user. be trained, so that incorrect classifications in part of the terrain can be easily corrected.

4 BE2022/5191

A preferred form of the unmanned autonomous vehicle is described in subclaim 17.

In a third aspect, the present invention concerns a use according to claim 18.

This use results in economical autonomous maintenance of a garden using an unmanned autonomous vehicle because the unmanned autonomous vehicle can immediately be used by a user for garden maintenance, without the user having to connect a neural network to the unmanned autonomous vehicle. and because the user can train a local network with very limited effort, if the unmanned autonomous vehicle performs an unwanted action due to incorrect classifications in a part of the garden or does not perform a desired action, in order to correct the incorrect classifications.

DESCRIPTION OF THE FIGURES

Figure 1 shows a schematic representation of a site, indicating different zones.

Figure 2 shows a schematic representation of global neural networks and local neural networks included in an unmanned autonomous vehicle, according to an embodiment of the present invention.

Figure 3 shows a schematic representation of a selection of neural networks and a set of rules depending on a position of an unmanned autonomous vehicle, according to an embodiment of the present invention, in a terrain.

DETAILED DESCRIPTION

Unless otherwise defined, all terms used in the description of the invention, including technical and scientific terms, have the meanings commonly understood by those skilled in the art of the invention. For a better appreciation of the description of the invention, the following terms are explicitly explained.

“A”, “the” and “it” in this document refer to both the singular and the plural unless the context clearly suggests otherwise. For example, “a segment” means one or more than one segment. 5 The terms “include”, “comprising”, “consist of”, “consisting of”, “providing”, “containing”, “containing”, “containing”, “containing” are synonyms and are inclusive or open terms which indicate the presence of what follows, and which do not exclude or preclude the presence of other components, features, elements, members, steps, known or described in the prior art.

Quoting numerical intervals through the endpoints includes all integers, fractions, and/or real numbers between the endpoints, including these endpoints.

In the context of this document, a neural network means an artificial neural network, where the neural network includes inputs, nodes called neurons, and outputs. An input is connected to one or more neurons. An output is also connected to one or more neurons. A neuron can be connected to one or more neurons. A neural network can include one or more layers of neurons between an input and an output. Each neuron and each connection of a neural network typically has a weight that is adjusted during a training phase using a training set of sample data.

In a first aspect, the invention concerns a method for controlling an unmanned autonomous vehicle based on a soil type on a site.

According to a preferred embodiment, the unmanned autonomous vehicle comprises a drive unit for moving the unmanned vehicle over the terrain, a camera for capturing images of the terrain, a positioning means for determining a position of the unmanned vehicle on the terrain. , a processor and memory and a tool.

The drive unit preferably comprises at least one wheel and a motor for driving the wheel. Preferably the motor is an electric motor. Preferably, the unmanned autonomous vehicle comprises a battery for powering the engine and other electrical systems. It is clear to a professional trained in the technical field that the unmanned autonomous vehicle has two, three, four or more wheels

6 BE2022/5191, whereby at least one wheel, preferably at least two wheels, are coupled to the motor for drive. It is clear to a skilled person trained in the technical field that the at least one wheel can be part of a caterpillar track, wherein the caterpillar track can be driven by the motor by means of the at least one wheel. The unmanned autonomous vehicle includes a steering device for steering the unmanned autonomous vehicle. The steering device is a conventional steering device in which at least one wheel is arranged to be rotatable. Alternatively, the steering device is part of the drive unit, where two wheels on opposite sides of the unmanned autonomous vehicle can be driven differently by the engine. Different means with different speed and/or opposite direction of rotation. The steering gear may or may not be part of the drive unit.

The camera is a digital camera. The camera is at least suitable for making two-dimensional images. The camera is optionally suitable for making three-dimensional images, with or without depth determination. The camera has a known viewing angle. The camera has a known position and alignment on the unmanned autonomous vehicle. The camera is aimed in such a way that at least part of the bottom of the site is captured in the image. Because the viewing angle of the camera and the position and alignment of the camera on the unmanned autonomous vehicle are known, a position, relative to the position of the unmanned vehicle, of a soil type of the terrain shown in an image is that is recorded, visible, known using the digital camera. The camera has a fixed position and alignment on the unmanned autonomous vehicle. Alternatively, the camera is arranged to be rotatable, whereby the camera can be rotated 360° in a horizontal plane and 180° rotatable in a vertical plane. The rotatable arrangement of the camera is preferably drivably linked to motors with encoders. Motors with encoders are advantageous for knowing the position and alignment of a rotatably mounted camera. Optionally, the camera is also suitable for taking images with non-visible light, such as infrared light or ultraviolet light. This is advantageous because it allows images of the terrain to be captured using visible light, infrared light and ultraviolet light, from which various information can be obtained, which can be advantageously combined for a successful classification of a soil type of the terrain visible in a captured image. It is clear to a professional trained in the technical field that instead of a single camera, different cameras can also be combined, where, for example, a first camera captures images using visible

7 BE2022/5191 light, a second camera captures images using infrared light and a third camera captures images using ultraviolet light. Preferably, the first camera, the second camera and the third camera have an overlapping field of view. This is advantageous for combining information from images captured using the first camera, the second camera and the third camera. It is clear to a professional trained in the technical field that the unmanned autonomous vehicle can include several similar cameras.

The positioning means for determining the position of the unmanned vehicle can be any suitable means. The positioning means is, for example, a Global Navigation Satellite System (GNSS), such as GPS, GLONASS or Galileo. The positioning means is, for example, a system with wireless beacons on the site, whereby the unmanned autonomous vehicle determines a position on the site by triangulation. The positioning means is, for example, based on recognition of reference points in images of the terrain, for example images made using the camera of the unmanned autonomous vehicle. Because the viewing angle of the camera and the position and alignment of the camera on the unmanned autonomous vehicle are known, it is possible, using trigonometry and/or photogrammetry, to determine a distance from a reference point on an image to the camera and the unmanned autonomous vehicle, to automatically estimate a distance between two reference points on an image and/or to automatically estimate a size of a reference point on an image, even if the camera is only suitable for making two-dimensional images, which allows the position of the unmanned autonomous vehicle on site can be determined.

The memory comprises a working memory and a non-volatile memory for storing, for example, data, such as images captured by the camera, and, for example, a program that is executed by the processor, whereby upon execution of the program the method is carried out.

The implement is any implement that is suitable for performing a task in the field. Non-exhaustive examples of suitable tools are a lawn mower, a vacuum cleaner, a brush, a spray lance, pruning shears, etc.

The method includes the steps of:

8 BE2022/5191 - Determining a position of the unmanned autonomous vehicle on the site using the positioning device. The site can be either indoors, for example a factory hall or a party hall, or outdoors, for example a garden, a football field or a square. - Capturing at least one image of the terrain using the camera of the unmanned autonomous vehicle. The at least one image may or may not be stored in the memory of the unmanned autonomous vehicle. The at least one image may or may not be forwarded by the unmanned autonomous vehicle to an external processing unit. It is clear to a professional trained in the technical field that if the unmanned autonomous vehicle moves to another position on the terrain, at least one image of the terrain is captured again using the camera of the unmanned autonomous vehicle. . Preferably, during movements by the unmanned autonomous vehicle along a route from a first position to a second position, positions are recorded at regular intervals. The images are captured at an adjusted speed, expressed in images per second, so that successive images overlap at least partially. It is clear to a professional trained in the technical field that the speed of capturing images depends on the speed at which the unmanned autonomous vehicle moves on the terrain. - Processing of at least one image of the terrain by the processor. If the camera has captured multiple images, the multiple images are preferably processed. The unmanned autonomous vehicle includes at least one global neural network and at least one local neural network.

A global neural network was trained using a global training set.

A global training set includes at least 70% images of terrains other than the terrain on which the unmanned autonomous vehicle is located. Preferably, a global training set includes at least 80% images from other areas, more preferably at least 90%, even more preferably at least 95%, and even more preferably at least 99.99%. A global training set includes images from many areas in order to obtain a global neural network that can be used as standard for a multitude of different areas. A global neural network is preferably trained on an external processing unit and in memory

9 BE2022/5191 of the unmanned autonomous vehicle. It is clear to a skilled person in the technical field that if the unmanned autonomous vehicle includes multiple global neural networks, the global neural networks are preferably trained with different global training sets. This does not mean that different global training sets cannot have images in common. Having at least one global neural network is advantageous because it allows the unmanned vehicle to be deployed immediately, without a user having to first create a training set of images of the entire terrain and train a neural network. This would require a lot of time from the user to capture a large number of images of the terrain and to annotate all these images.

This is also expensive for a user because a lot of memory storage is required to store the training set.

A local neural network was trained using a local training set. A local training set only includes images of the terrain where the unmanned autonomous vehicle is located. The images of a local training set were taken using a separate digital camera device, a GSM device with camera, a tablet with camera, the camera of the unmanned autonomous vehicle or another suitable camera and added to the said local training set. A local neural network was trained on an external processing unit and loaded into the memory of the unmanned autonomous vehicle. Alternatively, a local neural network has been trained on the processor of the unmanned autonomous vehicle. In that case, a local training set is also available in the memory of the unmanned autonomous vehicle.

Based on the position of the unmanned autonomous vehicle, which was determined in an earlier step of the method, one or more neural networks are selected from a group formed by the at least one global neural network and the at least one local neural network. In concrete terms, this means that one or more neural networks are selected from all global neural networks and all local neural networks included in the unmanned autonomous vehicle.

During processing of the at least one image, each of the selected neural networks determine at least one classification of at least one soil type in the at least one image. A classification can be binary, for example a bo-

10 BE2022/5191 soil type is grass (1) or is not grass (0), or can also be a value that represents a probability, for example a soil type with an 83% chance is grass. In the context of this document, a positive classification means a classification with a probability greater than 60%, preferably greater than 75%, more preferably greater than 90% and even more preferably greater than 98%. It is clear to a person skilled in the technical field that in a binary system a value of 1 is a positive classification and a value of 0 is not a positive classification. A selected neural network can be trained for one or more classifications. Non-exhaustive examples of classifications are grass, gravel, stone floor, soil, flower bed, leaves, parquet, vegetable garden, etc. Selected neural networks can be trained for different classifications, equal classifications or for only partially different or equal classifications. An image can include a single soil type. An image can contain multiple soil types, which are separated from each other, run through each other or a combination of both.

The at least one local neural network is advantageous to obtain a better classification using the at least one local neural network in specific cases where determining at least one classification of at least one soil type in an image by the global network results in an incorrect classification.

The at least one local neural network is additionally advantageous because only a lower number of images in the local training set is required to obtain a better classification in those specific cases, compared to retraining the at least one global neural network by adding images of the terrain to the global training set. The latter solution would require a huge number of images to have any impact on the training of at least one global neural network for those specific cases, as the global training set already includes a plethora of images from a large number of different areas. It is particularly advantageous that one or more neural networks are selected from the group formed by the at least one global neural network and the at least one local neural network based on the position of the unmanned autonomous vehicle. This means that a user does not have to train a local neural network for the entire terrain, but only for those parts where a global neural network produces an incorrect classification, which necessitates capturing the images for a local training set and training a local neural network. network only requires a limited amount of time and limited memory. Because there are multiple global neural networks and/or multiple local neural networks

11 BE2022/5191 networks can be selected, it is possible to make an optimal selection based on the desired classifications in a part of the site.

For example, a large part of a garden could be a lawn. For images of the lawn, a correct grass classification is determined everywhere in the garden, except around a tree. There, due to the shade and because the tree absorbs a lot of water, the grass is much scarcer and soil is visible through the grass, which means that a global neural network does not label it as grass. The unmanned autonomous vehicle would therefore, for example, not cut the grass at those positions. If the global neural network has to be retrained to also regard this part of the garden as grass, an enormous number of images of the lawn around the tree must be added to the global training set, which is practically not feasible. In addition, there is the real risk that a flower bed with some wild grass at the edge of the lawn will also be labeled as grass, causing the unmanned autonomous vehicle to also drive away from the flower bed, which is not desirable. By training a local neural network with a limited set of images and selecting this local neural network only around the tree, the sparse grass around the tree is removed correctly. If there is also a flowerbed in the vicinity of the tree, a second local neural network can be trained with images of the flowerbed and, for example, both local neural networks can be selected in the vicinity of the tree to be successful. to distinguish between the sparse grass and the flower bed around the tree. - Creation of at least one control signal by the processor for the drive unit and/or the implement. The at least one control signal is created on the basis of the classifications determined by the selected neural networks. The at least one control signal determines which action the drive unit and/or the implement performs. For example, at least one control signal determines how the unmanned autonomous vehicle moves or which tool is used. This is advantageous because a user of an unmanned autonomous vehicle does not have to determine for each position on the site which action the unmanned autonomous vehicle will perform. It is sufficient to determine which action should be carried out for a particular soil type. For example, if a stone floor is determined as a positive classification for the soil type, the at least one control signal determines that the unmanned autonomous vehicle drives further over the stone floor and sweeps the stone floor. Other examples include cutting grass if grass has a positive classification for the soil type

12 BE2022/5191 is determined, watering a flower bed if flowers are determined as a positive classification for the soil type, sweeping up leaves if fallen leaves is determined as a positive classification for the soil type, ...

If an image includes multiple soil types that are separated from each other, the at least one control signal is preferably created based on a classification for a soil type immediately adjacent to the unmanned autonomous vehicle, preferably in a direction of travel of the unmanned autonomous vehicle. If an image includes multiple soil types that intermingle, the at least one control signal is preferably created based on multiple classifications for the soil types immediately adjacent to the unmanned autonomous vehicle, preferably in a direction of travel of the unmanned autonomous vehicle. For example, for an area immediately adjacent to the unmanned autonomous vehicle, a histogram of soil type classifications is created, with at least one control signal created based on a classification with a highest value in the histogram.

According to a preferred embodiment, the at least one control signal is created according to predetermined rules on the basis of the classifications determined by the selected neural networks. At least two different sets of predetermined rules are defined. Depending on the position of the unmanned autonomous vehicle, a set of predetermined rules is selected.

This embodiment is advantageous for determining other actions by the unmanned autonomous vehicle depending on the position of the unmanned autonomous vehicle on the terrain. For example, fallen leaves may lie on a lawn, meaning that grass may no longer be visible in certain parts of the lawn. The unmanned autonomous vehicle could mow the grass even further, at least partially shredding the fallen leaves. For example, at the edge of the lawn, a border stone may be hidden under the fallen leaves. If the unmanned autonomous vehicle drives further down the grass here, it is very likely that the unmanned vehicle's lawn mowing tool will be damaged by the coping stone. It is therefore not desirable to mow the grass further here. Based on the classifications determined by the selected neural networks, it is not possible to distinguish between the two positions. For both positions in the terrain it is possible that only fallen leaves were determined as a positive classification for the soil type. By now for it

13 BE2022/5191 lawn to select a first set of rules, which, for example, says that mowing is allowed in both grass and fallen leaves, and near the edge of the lawn to select a second set of rules, which, for example, says that only grass is allowed mowing, and not on fallen leaves, can avoid creating a control signal that causes the unmanned autonomous vehicle to mow over the curb and damage the mowing implement.

This embodiment is additionally advantageous if an image comprises several soil types that overlap each other. The at least one control signal is created on the basis of several positive classifications for the soil types and on the basis of the selected set of rules. For example, a rule may stipulate that if both grass, soil and weeds have been determined as positive classifications for soil types, mowing takes place at these positions. This is particularly advantageous compared to a control signal created on the basis of a classification with a highest value in a histogram, as in a previously described embodiment. For example, in a severe drought it is possible that grass is much scarcer than usual and there is more soil and weeds, causing soil to be the highest value in the histogram, and a decision could be made not to mow while the unmanned autonomous vehicle moves around. is indeed on a lawn.

An advanced option is that at least one control signal is created based on multiple positive classifications for the soil types and based on the selected set of rules, where the selected set of rules sets a threshold value for an area of at least one soil type with a positive rating. In the example above where grass, soil, and weeds were determined as positive classifications for soil types, the selected set of rules may include a minimum grass area value. This threshold value can be used as an absolute value, for example an area in m2 or a number of pixels in a camera image, or as a relative value, for example a percentage of a complete area in m2 or in pixels in a camera image or a share in a histogram. , are expressed. In the example above, the selected set of rules would have a threshold value of, for example, at least 50% grass or at least 3 m? may include grass to be allowed to mow in that position. The selected set of rules could include a threshold value that, for example, is at most 30% soil or at most 0.5 m? soil allows mowing at that position. It is also clear from this example that threshold values can be logically combined.

14 BE2022/5191

According to a further embodiment, a specific selection is made from a set of predetermined rules for each control signal based on the position of the unmanned autonomous vehicle.

This embodiment is advantageous because it allows a new set of predetermined rules to be selected at a certain position for a first control signal, while for a second control signal an already selected set of rules can be retained or, for example, another set of rules can be retained. predetermined rules can be selected. For example, for a control signal for the drive unit at a position where the terrain is often muddy, a new set of predetermined rules could be selected, reducing the speed of the unmanned autonomous vehicle if soil was determined as positive classifications for the soil type, in order to to avoid drive unit wheels from slipping and damaging the terrain, while the control signal for a fallen leaf picking tool still uses the same set of predetermined rules.

According to an embodiment, the one or more neural networks from the group formed by the at least one global neural network and the at least one local neural network are also selected on the basis of time in addition to the position of the unmanned autonomous vehicle. For example, time can be a time of a day, a day of a week, a month, or a season. This embodiment is advantageous because a site may have a different appearance during a day, week, month or season, making correct classification of soil types difficult. For example, the site may be in the shade at dusk, certain types of flowers may only be present in the spring, etc. By selecting other neural networks, the selected neural networks can be optimized for a time. This embodiment applies to both global neural networks and local neural networks.

According to one embodiment, the set of predetermined rules in addition to the position of the unmanned autonomous vehicle is also selected based on time. For example, time can be a time of a day, a day of a week, a month, or a season. This embodiment is advantageous because other tasks are necessary on a site during a day, week, month or season. For example, certain plants are preferably watered in the morning or evening. For example, drought is less of a problem in seasons other than summer than in summer

15 BE2022/5191 summer. In the previously described example of mowing grass, outside of summer a different set of rules would be selected so that if both grass, soil and weeds were determined as positive classifications for soil types, these positions would not be mowed instead of mowed, because it is unlikely that the unmanned autonomous vehicle is still on the lawn.

According to a preferred embodiment, a specific selection of one or more neural networks from the group formed by the at least one global neural network and/or the at least one local neural network is made for each control signal based on the position of the unmanned autonomous vehicle.

This embodiment is advantageous because it allows a new selection of neural networks to be made at a certain position for a first control signal, while for a second control signal an already made selection of neural networks can be retained or, for example, another selection of neural networks can be made. can be made.

For example, for a control signal for a mowing implement at a position on the site where there are also flower beds in the grass, a new selection of neural networks could be made, where, for example, a local neural network is added to the selection of neural networks to use the flower beds as to be able to classify soil type, to prevent the mowing implement from mowing the flower beds, while the control signal for the drive still uses the same selection of neural networks.

According to a preferred embodiment, a first group of zones are defined on a digital map of the site. Preferably, the digital map is presented visually in a graphics application, whereby the first group of zones are drawn graphically on the digital map. The graphic application is preferably suitable for use on a smartphone and/or a tablet and/or a computer. The graphical application is preferably suitable for use in a web browser. Associated with each zone of the first group of zones is a specific selection of one or more neural networks from the group formed by the at least one global neural network and/or the at least one local neural network. Preferably, the association also takes place in the aforementioned graphic application. The digital map of the terrain includes the first group of zones and the associated selections of one or more neural networks from the group formed by the

16 BE2022/5191 at least one global neural network and/or at least one local neural network.

The digital map is loaded into the unmanned autonomous vehicle. The digital map is loaded into the unmanned vehicle using a data cable. A non-exhaustive example is a USB cable. Alternatively, the digital map is loaded into the unmanned vehicle via a wireless connection. Non-exhaustive examples are a Bluetooth connection or a WiFi connection. The unmanned autonomous vehicle determines the zone from the first group of zones in which the unmanned autonomous vehicle is located based on the position of the unmanned autonomous vehicle on the site. The unmanned autonomous vehicle uses the digital map for this. The unmanned autonomous vehicle uses the associated selection of one or more neural networks from the group formed by the at least one global neural network and/or the at least one local neural network while processing the at least one image of the terrain.

This embodiment is advantageous because a user can easily define on a digital map a first group of zones on the site where, for example, the use of a standard global neural network for correct classification is not sufficient and where a different selection of neural networks is required. By using a digital map, this does not require intervention from, for example, a technician. The user only needs to associate an appropriate selection of neural networks with a zone from the first group of zones. This can be done easily in a graphic application, for example.

According to a preferred embodiment, a second group of zones are defined on a digital map of the site. The digital map may or may not be the same as the digital map in a previously described embodiment on which a first group of zones is defined. The digital map can preferably be presented in a similar manner as in the aforementioned previously described embodiment and the second group of zones can preferably be defined in a similar manner. Each zone from the second group of zones is associated with a specific selection from a set of predetermined rules. Preferably the association also takes place as in the aforementioned previously described embodiment in a graphic application. The digital map includes the zones from the second group of zones and the associated selections of sets of predetermined rules. The digital map is loaded into the unmanned autonomous vehicle.

Preferably, the digital card is loaded into the unmanned autonomous vehicle in a similar manner as in the previously described embodiment. It

17 BE2022/5191 unmanned autonomous vehicle determines the zone from the second group of zones where the unmanned autonomous vehicle is located based on the position of the unmanned autonomous vehicle on the terrain. The unmanned autonomous vehicle uses the associated set of predetermined rules while creating the at least one control signal.

This embodiment is advantageous because a user can easily define a second group of zones on the site on a digital map where, for example, a different set of predetermined rules is required, for example because with equal positive classifications of a soil type, a different action is required due to the mande autonomous vehicle is desired. By using a digital card, this does not require intervention from, for example, a technician. The user only needs to associate an appropriate set of predetermined rules with a zone from the second group of zones. This can be done easily in a graphic application, for example.

According to a further embodiment, the first group of zones is equal to the second group of zones. This is advantageous because it makes it clear to a user which selection of neural networks and which set of predetermined rules will be used by the unmanned autonomous vehicle at a position in the terrain. The zones only need to be defined once.

According to a preferred embodiment, zones are defined hierarchically. Associated selections of neural networks and/or associated sets of predetermined rules from a hierarchically higher zone are automatically associated with a hierarchically lower zone. Associated selections of neural networks and/or associated sets of predetermined rules from a hierarchically lower zone are not associated with a hierarchically higher zone. Preferably, an associated selection of neural networks and/or associated set of predetermined rules from a lower hierarchical zone for a specific control signal takes precedence over an associated selection of neural networks and/or associated set of predetermined rules from a higher hierarchical zone. re hierarchical zone.

This embodiment is advantageous for easily defining smaller zones where at least one control signal must be created in a different way, for example because a different or additional action by the unmanned autonomous

18 BE2022/5191 vehicle must be carried out or, for example, because a different selection of neural networks is required to obtain correct classifications for a particular action, so that a correct control signal for the said action is effectively created. For example, an action can be mowing the grass, but it can also be stopping mowing the grass. For other control signals, the associated selection of neural networks and/or the associated set of predetermined rules from a hierarchically higher zone can still be used. This eliminates the need to divide the entire site into many smaller zones and to associate a selection of neural networks and/or a set of predetermined rules with all zones for all possible control signals.

According to a preferred embodiment, images for a local training set are captured using the camera of the unmanned autonomous vehicle. This embodiment is advantageous because no additional camera is required to capture images for the local training set. This embodiment is additionally advantageous because images for the local training set are captured in the same orientation relative to the unmanned autonomous vehicle as when using the unmanned autonomous vehicle to perform tasks. This embodiment is particularly advantageous because the images for the local training set can be captured while using the unmanned autonomous vehicle to perform tasks or during an autonomous exploration of at least part of the terrain.

According to a further embodiment, the images of the local training set are processed for training a local neural network using the processor of the unmanned autonomous vehicle. The images of the local training set are stored in the memory of the unmanned autonomous vehicle.

The images are labeled using an external device, such as a smartphone, a tablet or a computer, or on a screen included in the unmanned autonomous vehicle. In case the images are tagged on an external device, the images are first sent from the unmanned vehicle over a wired or wireless connection to the external device, after which the labels are sent back to the unmanned autonomous vehicle. Non-limiting examples of suitable wired or wireless connections are given in previously described embodiments. The labeled images are processed by the processor to train the local neural network.

19 BE2022/5191

This embodiment is advantageous because no capacity on external processing units, such as a server in a cloud environment, needs to be reserved to train a local neural network.

According to a preferred embodiment, the images of the local training set, which were captured using the camera of the unmanned autonomous vehicle, are forwarded to an external processing unit and added to one or more global training sets. The images can be labeled before or after. Preferably, the images are labeled afterwards, preferably by a technician. Optionally, a selection of images from the local training set is added to one or more global training sets. The selection of images from the local training set does not have to be the same for every global training set. After adding the images of the local training set to one or more global training sets, the one or more global training sets are processed by the external processing unit for training one or more global neural networks. This embodiment is advantageous for incrementally obtaining large global training sets and for incrementally improving global neural networks. This is particularly advantageous because local training sets are often created in case global neural networks do not classify soil types with sufficient accuracy. By incorporating images from local training sets into global training sets, future users of other unmanned autonomous vehicles can avoid similar problems.

According to a preferred embodiment, the unmanned autonomous vehicle determines its position using a digital map of the terrain and on the basis of images of the terrain, which are captured using the camera of the unmanned autonomous vehicle. Reference points in the images are compared with reference points on the digital map. This embodiment is advantageous because it means that an unmanned autonomous vehicle only needs a camera for both capturing images for classification of soil types and for determining the position of the unmanned autonomous vehicle on the ground.

According to a preferred embodiment, at least one of the selected neural networks determines at least one classification of at least one object in the at least one image. This can be either a global neural network and/or a local neural network. The at least one control signal is generated on the basis of the

20 BE2022/5191 selected neural networks created certain classifications for soil types and objects. Previously described embodiments are also applicable mutatis mutandis for training a local or neural network or a global neural network for classifying an object. This embodiment is advantageous if a task to be performed by the unmanned autonomous vehicle depends not only on a soil type, but also on the presence or absence of an object, for example the presence of a charging station for charging a battery. of the unmanned autonomous vehicle or the availability of a waste container for dumping household or garden waste, for example.

According to a preferred embodiment, each of the classifications determined by the selected neural networks is assigned a weight. The weight depends on the position of the unmanned autonomous vehicle. For example, weights are advantageous for use as threshold values in predetermined rules. Weights are, for example, advantageous if an image contains multiple soil types that intermingle to give more or less weight to a certain positive classification. Weights are particularly advantageous for use at the edges of zones. Weights that change gradually as they approach the edge of a zone can provide a smooth transition from a first zone to a second zone, for example by gradually transitioning weights that correspond to a first predetermined rule to weights in a first zone corresponding to a second predetermined rule in a second zone, or for example by gradually transferring weights given to certain positive classifications in a first zone to weights given to the same positive classifications in a second zone.

According to a preferred embodiment, it is determined on the basis of the classifications determined by the selected neural networks whether a soil type belongs to a set of predetermined soil types. The unmanned autonomous vehicle only moves over soil types from the collection of predetermined soil types.

This embodiment is advantageous because it allows the unmanned autonomous vehicle to move autonomously across a site, whereby the unmanned autonomous vehicle remains within a perimeter on the site. The perimeter is determined by a transition between soil types that belong to the set of predetermined soil types and soil types that do not belong to the set of predetermined soil types.

21 BE2022/5191 soil types. The unmanned autonomous vehicle cannot exceed this transition. For example, grass belongs to the set of predetermined soil types and flower beds, soil and terrace do not belong to the set of predetermined soil types. This keeps the unmanned autonomous vehicle on a lawn bordered by flower beds and a terrace. It is therefore not necessary to create physical boundaries around the lawn or to place or bury a signal wire around part of the site.

According to a further embodiment, the set of predetermined soil types depends on the position of the unmanned autonomous vehicle on the site. This embodiment is particularly advantageous for creating corridors for the unmanned vehicle between different zones on the site. For example, there are two tiled terraces on the site that are separated from each other by a gravel path. The unmanned autonomous vehicle, for example, includes a brush as a tool for brushing the tiled terraces. The predetermined collection of soil types consists of tiled terraces, so the unmanned vehicle will only move over the tiled terraces. This does mean that the unmanned autonomous vehicle will only brush one tiled terrace, because the unmanned vehicle cannot cross the gravel path to move to the other tiled terrace. Depending on the position of the unmanned autonomous vehicle on the site, a gravel path can be added as a soil type to the set of predetermined soil types at the height of a desired corridor, so that the unmanned autonomous vehicle at the corridor can move over the gravel path to the another tiled terrace can move. At positions outside the corridor, a gravel path has not been added to the set of predetermined soil types, so the unmanned autonomous vehicle does not enter the gravel path there. It is clear that in this example, based on the position of the unmanned autonomous vehicle, it can be ensured that a control signal for the brush is created so that the unmanned autonomous vehicle does not brush the gravel path.

In a second aspect, the invention concerns an unmanned autonomous vehicle for performing tasks on a site.

According to a preferred embodiment, the unmanned autonomous vehicle comprises a drive unit for moving the unmanned vehicle over

22 BE2022/5191 the terrain, a camera for capturing images of the terrain, a positioning device for determining a position of the unmanned vehicle on the terrain and a tool.

The drive unit preferably comprises at least one wheel and a motor for driving the wheel. Preferably the motor is an electric motor. Preferably, the unmanned autonomous vehicle comprises a battery for powering the engine and other electrical systems. It is clear to a person skilled in the technical field that the unmanned autonomous vehicle can comprise two, three, four or more wheels, with at least one wheel, preferably at least two wheels, being coupled to the engine for propulsion. It is clear to a skilled person trained in the technical field that the at least one wheel can be part of a caterpillar track, wherein the caterpillar track can be driven by the motor by means of the at least one wheel. The unmanned autonomous vehicle includes a steering device for steering the unmanned autonomous vehicle. The steering device is a conventional steering device in which at least one wheel is arranged to be rotatable. Alternatively, the steering device is part of the drive unit, where two wheels on opposite sides of the unmanned autonomous vehicle can be driven differently by the engine. Different means with different speed and/or opposite direction of rotation. The steering gear may or may not be part of the drive unit.

The camera is a digital camera. The camera is at least suitable for making two-dimensional images. The camera is optionally suitable for making three-dimensional images, with or without depth determination. The camera has a known viewing angle. The camera has a known position and alignment on the unmanned autonomous vehicle. The camera is aimed in such a way that at least part of the bottom of the site is captured in the image. Because the viewing angle of the camera and the position and alignment of the camera on the unmanned autonomous vehicle are known, a position, relative to the position of the unmanned vehicle, of a soil type of the terrain shown in an image is that is recorded, visible, known using the digital camera. The camera has a fixed position and alignment on the unmanned autonomous vehicle. Alternatively, the camera is arranged to be rotatable, whereby the camera can be rotated 360° in a horizontal plane and 180° rotatable in a vertical plane. The rotatable arrangement of the camera is preferably drivably linked to motors with encoders. Motors with encoders are advantageous for knowing the position and alignment of a rotary

23 BE2022/5191 camera installed. Optionally, the camera is also suitable for taking images with non-visible light, such as infrared light or ultraviolet light. This is advantageous because it allows images of the terrain to be captured with visible light, infrared light and ultraviolet light, from which various information can be obtained. It is clear to a professional trained in the technical field that instead of a single camera, different cameras can also be combined, where, for example, a first camera images using visible light, a second camera images using infrared. light and a third camera capture images using ultraviolet light. Preferably, the first camera, the second camera and the third camera have an overlapping field of view. This is advantageous for combining information from images captured using the first camera, the second camera and the third camera. It is clear to a professional trained in the technical field that the unmanned autonomous vehicle can include several similar cameras.

The positioning means for determining the position of the unmanned vehicle can be any suitable means. The positioning means is, for example, a Global Navigation Satellite System (GNSS), such as GPS, GLONASS or Galileo. The positioning means is, for example, a system with wireless beacons on the site, whereby the unmanned autonomous vehicle determines a position on the site by triangulation. The positioning means is, for example, based on recognition of reference points in images of the terrain, for example images made using the camera of the unmanned autonomous vehicle. Because the viewing angle of the camera and the position and alignment of the camera on the unmanned autonomous vehicle are known, it is possible, using trigonometry and/or photogrammetry, to determine a distance from a reference point on an image to the camera and the unmanned autonomous vehicle, to automatically estimate a distance between two reference points on an image and/or to automatically estimate a size of a reference point on an image, even if the camera is only suitable for making two-dimensional images, which allows the position of the unmanned autonomous vehicle on site can be determined.

The implement is any implement that is suitable for performing a task in the field. Non-exhaustive examples of suitable tools are a lawn mower, a vacuum cleaner, a brush, a spray lance, pruning shears, etc.

24 BE2022/5191

The unmanned autonomous vehicle includes a memory and a processor. The processor is configured to perform a method according to the first aspect. Memory includes working memory and non-volatile memory.

Such an unmanned autonomous vehicle is advantageous because it can be immediately deployed on a site to perform tasks without additional training of a neural network by a user of the unmanned autonomous vehicle, while a local neural network can be used with limited effort by the user. be trained, so that incorrect classifications in part of the terrain can be easily corrected.

According to a preferred embodiment, the camera of the unmanned vehicle is only suitable for making two-dimensional images. Preferably, only one camera for two-dimensional images is mounted on the unmanned autonomous vehicle.

This embodiment is particularly advantageous because it results in a very simple unmanned autonomous vehicle for carrying out tasks on a site.

A skilled person trained in the technical field will appreciate that a method according to the first aspect is preferably carried out with an unmanned autonomous vehicle according to the second aspect and that an unmanned autonomous vehicle according to the second aspect is preferably configured for carrying out a method. according to the first aspect. Each feature described in this document, above as well as below, may therefore relate to any of the three aspects of the present invention.

In a third aspect, the invention concerns a use of a method according to the first aspect and/or an unmanned autonomous vehicle according to the second aspect for autonomously maintaining a garden.

This use results in economical autonomous maintenance of a garden using an unmanned autonomous vehicle because the unmanned autonomous vehicle can immediately be used by a user for garden maintenance, without the user having to connect a neural network to the unmanned autonomous vehicle. and because the user can train a local network with very limited effort, if the unmanned autonomous vehicle is damaged by incorrect classifications in

25 BE2022/5191 part of the garden performs an undesirable action or does not perform a desired action, to correct the incorrect classifications.

In what follows, the invention is described by means of: non-limiting figures illustrating the invention, and which are not intended or should be interpreted to limit the scope of the invention.

FIGURE DESCRIPTION

Figure 1 shows a schematic representation of a site, indicating different zones.

The site includes seven zones. The zones are hierarchically divided into two layers.

A highest hierarchical layer comprises a first zone (1), which is shaded in Figure 1, and within zone (1) three adjacent zones, namely a second zone (2), a third zone (3) and a fourth zone ( 4). A lowest hierarchical layer comprises three zones, namely a fifth zone (5) located within the second zone (2), a sixth zone (6) located within the third zone (3) and a seventh zone (7) located is located within the fourth zone (4).

In this example, the site is a garden, where the first zone (1) is for example a border with shrubs, the second zone (2) is a lawn, the third zone (3) is a lawn with a concrete path and the fourth zone (4) ) again a lawn. A charging station has been installed in zone (5). A compost bin has been placed in zone (6).

The zone (7) is a sandbox.

Figure 2 shows a schematic representation of global neural networks and local neural networks included in an unmanned autonomous vehicle, according to an embodiment of the present invention.

In this example, the unmanned autonomous vehicle has two global neural networks (A) and (B) and two local neural networks (C) and (D).

The global neural network (A) is trained to determine six classifications (a), (b), (c), (d), (e), and (f). In this example, classification (a) is grass, classification (b) clover, classification (c) soil, classification (d) dandelion, classification (e) tree bark and classification (f) a charging station.

26 BE2022/5191

The global neural network (B) is trained to determine five classifications (f), (9), (h), (i) and (j). The classifier (f) is again a loading station, but the global neural network (B) in this example has been trained with a different global training set, which means that the global neural networks (A) and (B) for classification (f) can obtain different results . Classification (g) is a person, classification (h) is a tree, classification (i) is a car and classification (j) is a plant. In this example, the global neural network (B) is specifically trained to determine classifications of objects instead of determining classifications of soil types.

The local neural network (C) is trained to determine two specific classifications for the specific terrain in this example. Classification (k) is a compost bin and classification (I) is a bucket. The local neural network (D) is trained to determine one specific classification for the specific terrain in this example. Classification (m) is sand.

Figure 3 shows a schematic representation of a selection of neural networks and a set of rules depending on a position of an unmanned autonomous vehicle, according to an embodiment of the present invention, in a terrain.

The terrain is the terrain from the example of Figure 1. The neural networks are the global neural networks (A) and (B) and the local neural networks (C) and (D) from Figure 2.

In this example, a single action (A1) is defined at the highest hierarchical level of zones, namely cutting grass. The global neural network (A) is associated with all zones of the highest hierarchical level (1), (2), (3) and (4). Associated with zone (1) is a set of predetermined rules (C1), which stipulate that the grass is only mowed if there are positive classifications for grass (a), for dandelions (d) and for tree bark (e). are. This means that the unmanned autonomous vehicle is located in the border between the bushes, where there is tree bark and where wild grass and dandelions grow, which can be mowed. The lawn in zone (2) was dug up and sown with clover just before redevelopment. Associated with zone (2) is a set of predetermined rules (C2), which stipulate that the grass will only be mown if there are positive classifications for clover (b) and for soil (c). This means that the unmanned autonomous vehicle is on the plowed lawn, where any

27 BE2022/5191 grass and clover may be mowed. Optionally, the set of predetermined rules (C2) can include a threshold value for soil (c), whereby mowing at a position in zone (2) is only allowed if a surface of the soil (c) is lower than the threshold value. This threshold value can be defined as a relative or as an absolute value. In zone (3) there is a concrete path. Associated with zone (3) is a set of predetermined rules (C3), which determine that the grass is only mowed if there is a positive classification for grass (a). This means that the unmanned autonomous vehicle is definitely on grass and not on the concrete path, which prevents the unmanned vehicle from being damaged by mowing on the concrete path. Associated with zone (4) is a set of predetermined rules (C4), which stipulate that the grass is only mowed if there are positive classifications for grass (a) and for dandelions (d). The intention is that the grass in zone (4) is longer and wilder.

By only mowing the grass where there are dandelions, the grass can grow until dandelions can grow.

In this example, at the lowest hierarchical level of zones, three actions (A2), (A3) and (A4) are defined, namely charging a battery of the unmanned autonomous vehicle (A2), dumping garden waste in a compost bin (A3 ) and sweeping a sandbox (A4). Both the global neural network (A) and the global neural network (B) are associated with zone (5). Associated with zone (5) is a set of predefined rules (C5), which determine that the battery of the unmanned autonomous vehicle may be charged if there is a positive classification for a charging station (f), both by the global neural network (A ) and the global neural network (B). Because the global neural network (A) and the global neural network (B) are trained with a different global training set, a reliable classification (f) is determined in this way. Both the global neural network (A) and the local neural network (C) are associated with zone (6). Associated with zone (6) is a set of predetermined rules (C6), which stipulate that garden waste may be dumped in the compost bin if positive classifications for grass (a), soil (c) and dandelion (d) are given by the global neural network (A) and a positive classification is for a compost bin (k) and not for a bucket (|) by the local neural network (C). The compost bin is placed in zone (6) in the grass next to the concrete path. Due to the placement of the compost bin, the grass around the compost bin is mowed less often, which means there are dandelions and the grass around the compost bin has disappeared here and there. By using the local neural network (C) it is possible to obtain a positive classification (k} for the

28 BE2022/5191 specific compost bin in the garden of the example, even if the compost bin is moved and the compost bin is not confused with a specific bucket that is also often used in the garden.

Only the local neural network (D) is associated with zone (7).

Associated with zone (7) is a set of predetermined rules (C6), which determine that the sandbox may be swept if there is a positive classification for sand (m).

Claims (18)

29 BE2022/5191 CONCLUSIONS 1. Method for controlling an unmanned autonomous vehicle based on a soil type on a terrain, wherein the unmanned autonomous vehicle has a drive unit for moving the unmanned vehicle over the terrain, a camera for capturing images of the terrain, a positioning means for determining a position of the unmanned vehicle on the terrain, a processor and memory, and a tool, comprising the steps of: - determining a position of the unmanned autonomous vehicle on the terrain using of the positioning means; - capturing at least one image of the terrain using the camera of the unmanned autonomous vehicle; - processing at least one image of the terrain by the processor; - creation of at least one control signal by the processor for the drive unit and/or the implement; characterized in that the unmanned vehicle comprises at least one global neural network and one local neural network, wherein a global neural network is trained using a global training set and a local neural network is trained using a local training set, wherein a local training set only contains images of the terrain and where a global training set contains at least 70% images of other terrains, where, based on the position of the unmanned autonomous vehicle, one or more neural networks from a group formed by at least one global neural network and the at least one local neural network are selected, wherein during the processing of the at least one image by the processor, each of the selected neural networks determines at least one classification of at least one soil type in the at least one image, and wherein the at least one control signal is created based on the classifications determined by the selected neural networks. Method according to claim 1, characterized in that the at least one control signal is created according to predetermined rules on the basis of the classifications determined by the selected neural networks, whereby at least two different sets of predetermined rules are divided into 30 BE2022/5191 and where a set of predetermined rules is selected depending on the position of the unmanned autonomous vehicle. Method according to claim 2, characterized in that for each control signal a specific selection is made from a set of predetermined rules on the basis of the position of the unmanned autonomous vehicle. 4. Method according to any of the preceding claims 1-3, characterized in that for each control signal based on the position of the unmanned autonomous vehicle, a specific selection of one or more neural networks from the group formed by the at least one global neural network and/or at least one local neural network is created. 5. Method according to any of the preceding claims 1-4, characterized in that a first group of zones is defined on a digital map of the terrain, with each zone from the first group of zones being a specific selection of one or more neural networks from the group formed by the at least one global neural network and/or the at least one local neural network is associated, whereby the digital map of the terrain, comprising the zones from the first group of zones and the associated se- readings of one or more neural networks from the group formed by the at least one global neural network and/or the at least one local neural network are loaded into the unmanned autonomous vehicle, and wherein the unmanned autonomous vehicle determines the position based on its position determines the zone from the first group of zones in which the unmanned autonomous vehicle is located and the associated selection of one or more neural networks from the group formed by the at least one global neural network and/or the at least one local neural network during the processing at least one image of the terrain used. 6. Method according to any of the preceding claims 2-5, characterized in that a second group of zones is defined on a digital map of the terrain, with each zone from the second group of zones being a specific selection of a collection of predetermined rules is associated with the digital map of the terrain, including the zones from the 31 BE2022/5191 second group of zones and the associated selections of sets of predetermined rules, is loaded into the unmanned autonomous vehicle, and whereby the unmanned autonomous vehicle determines, based on its position, the zone from the second group of zones in which the unmanned autonomous vehicle is located and uses the associated set of predetermined rules while creating the at least one control signal. 7. Method according to any of the preceding claims 5-6, characterized in that the first group of zones is equal to the second group of zones. Method according to one of the preceding claims 5-7, characterized in that zones are defined hierarchically, wherein associated selections of neural networks and/or associated sets of predetermined rules of a hierarchically higher zone are automatically assigned a hierarchical lower zone and wherein associated selections of neural networks and/or associated sets of predetermined rules from a hierarchically lower zone are not associated with a hierarchically higher zone. 9. Method according to any of the preceding claims 1-8, characterized in that images for a local training set are recorded using the camera of the unmanned autonomous vehicle. Method according to claim 9, characterized in that the images of the local training set are processed for training a local neural network using the processor of the unmanned autonomous vehicle. Method according to one of the preceding claims 9-10, characterized in that the images of the local training set are forwarded to an external processing unit and added to one or more global training sets, after which the one or more global training sets are processed by the external processing unit for training one or more global neural networks. 12. Method according to any of the preceding claims 1-11, characterized in that the unmanned autonomous vehicle determines its position with the aid of a digital map of the terrain and on the basis of images of the terrain obtained with 32 BE2022/5191 using the camera of the unmanned autonomous vehicle, where reference points in the images are compared with reference points on the digital map. Method according to any one of the preceding claims 1-12, characterized in that at least one of the selected neural networks determines at least one classification of at least one object in the at least one image, wherein the at least one control signal is based on the data provided by the selected neural networks certain classifications for soil types and objects are created. Method according to any of the preceding claims 1-13, characterized in that a weight is assigned to each of the classifications determined by the selected neural networks, wherein the weight depends on the position of the unmanned autonomous vehicle. 15. Method according to any of the preceding claims 1-14, characterized in that on the basis of the classifications determined by the selected neural networks, it is determined whether a soil type belongs to a set of predetermined soil types, wherein the unmanned autonomous vehicle is only moved over soil types from the collection of predetermined soil types. 16. Unmanned autonomous vehicle for performing tasks on a site, comprising a drive unit for moving the unmanned vehicle over the site, a camera for capturing images of the site, a positioning means for determining a position of the unmanned vehicle in the field, a processor and memory, and a tool, characterized in that the processor is configured to carry out a method according to any one of claims 1-15. 17. Unmanned autonomous vehicle according to claim 16, characterized in that the camera of the unmanned autonomous vehicle is only suitable for making two-dimensional images. 33 BE2022/5191 Use of a method according to any one of claims 1-15 and/or an unmanned autonomous vehicle according to any one of claims 16-17 for autonomously maintaining a garden.