**4. Variable rate application (VRA) in postemergence (POST)**

The purpose of a POST application is to control weeds that have already emerged in the field. Thus, the target of the application is the aerial part of the plant species. For the VRA to be used in POST, it is necessary that the system has information about the weed population in the area. This information can be collected by the mapbased and sensor-based systems. Therefore, both methods can be used VRA in POST.

#### **4.1 Map based: weed mapping**

The literature mentions several methodologies for weed mapping, where each one has its specificity. Some have processing algorithms to differentiate monocot and eudicot plants [23]. Others use machine learning with deep neural network to identify weeds [24, 25]. However, all have the principle based on the quantitative and qualitative identification of the infested area, generation of the recommendation map, and integration with the VRA system.

Remote sensing is generally considered one of the most important technologies for precision agriculture. This technology can be used in weed mapping. Remote sensing can monitor many crops and vegetation parameters through images at various wavelengths. Images can be acquired by satellites, manned aircrafts, or unmanned aerial vehicles (UAV). However, satellite imagery is often not the best option because of the low spatial resolution of images acquired and the restrictions of the temporal resolutions as satellites are not always available to capture the necessary images [26]. Considering the use of manned aircrafts, usually it results in high costs, and many times, it is not possible to carry out multiple flights to obtain more than a few crop images. UAVs' ability to fly at a low altitude results in ultra-high spatial resolution images of the crops (i.e., a few centimeters). This significantly improves the performance of the monitoring systems. Furthermore, UAV-based monitoring systems have high temporal resolution as they can be used at the user's will. This enhances the flexibility of the image acquisition process [27]. In addition, UAVs are a lot simpler to use and also cheaper than manned aircrafts. Moreover,

*Pests, Weeds and Diseases in Agricultural Crop and Animal Husbandry Production*

the use of the same may have reduced efficiency in some situations.

similar soil conditions.

the maps.

equal regardless of the position in space.

Studies show that the electrical conductivity measured by contact sensor adequately reflects the variation in clay contents of the studied soil, being efficient to generate soil texture maps, including in no-tillage areas [21]. **Figure 6** shows a conductivity map elaborated with the data collected in VARIS 3100; the lowest conductivity values correlated with lower clay contents. However, for high clay contents, the model was less efficient. Thus, the mapping of electrical conductivity can be a useful tool in the design of more homogeneous areas, which present more

Considering that other factors such as moisture, salt concentration, and total carbon remain in the same conditions, soils with higher clay contents conduct more electricity than those with sandier texture. However, these factors may vary and affect the correlation between electrical conductivity and soil texture. Therefore, as the electrical conductivity method does not quantify the CEC and soil OM contents,

There are companies on the market that provide the VRA service for herbicides in PRE, one of which is APagri which has the HTV® method which consists of a process developed and patented for the application of herbicides in PRE at the varied rate based on maps (**Figure 7**), that considers the clay, OM, and CEC content of the soil [22]. The objective is to adjust the dose according to the soil ability to retain each type of herbicide so that the final concentration in the soil solution is

Due to technological limitations, there is still no VRA available on the market for PRE herbicides based on sensors that read, process, and apply the herbicide without the need for the generation of maps. One of the great challenges of this market is precisely to eliminate this stage, in view of the costs of generating

*Interpolated map of electrical conductivity measured with mobile contact measurement equipment. Source:* 

**186**

**Figure 6.**

*Machado et al. [21].*

they are more efficient than the ground systems as they can cover a large field in a short amount of time and in a non-destructive way, which is very important. UAVs can gather images and derive data from the whole field that can be used to generate a precise weed cover map depicting the spots where the herbicide are needed in different rates [28].

A variety of different types of sensors can be used in an agricultural UAV depending on the different vegetation parameters that should be monitored. The main sensors used that meet the limitations mentioned above are: visible light sensors, red, green, and blue (RGB) color model, multispectral sensors, hyperspectral sensors, and thermal sensors. RGB are relatively low cost compared to the other types and can acquire high resolution images, are easy to use and operate, and are lightweight [29]. In addition, the information acquired requires simple processing. However, they are inadequate for analyzing a lot of vegetation parameters that require spectral information in the non-visible spectrum. Thus, commonly are used with the other types of sensors.

Multispectral or hyperspectral imaging sensors can acquire information about the vegetation's spectral absorption and reflection on several bands. Spectral information can be significantly helpful in assessing a lot of biological and physical characteristics of the plants. This information is important to determinate which weed species are in the field [30]. Multispectral and hyperspectral sensors are frequently used, despite their higher costs. However, a drawback of these sensors arises from the fact that it is required to apply more complex preprocessing methods in order to extract useful information from the captured images. The preprocessing procedure of spectral images often contains the radiometric calibration, geometric correction, image fusion, and image enhancement. The main difference between multispectral and hyperspectral sensors is the number of bands (or channels) that each sensor can capture and the width of the bands. Multispectral sensors capture 5–12 channels, while hyperspectral images can usually capture hundreds or thousands of bands, but in a narrower bandwidth. Multispectral sensors are used much more frequently than hyperspectral sensors due to their lower cost, but hyperspectral technology appears to have a lot of potential and is considered the future trend for crop phenotyping research. Thermal infrared sensors capture information about the temperature of the objects and generate images displaying them based on this information and not their visible properties. This type of sensors is used for very specific applications (irrigation management). As a result, they are not frequently used in remotely piloted aircraft applications of UAV systems that focus on monitoring other characteristics of the crops [26–28].

UAVs can acquire information for various features of the cultivated field by using specialized sensors. However, as mentioned above, there is still no standardized workflow or well-established techniques to analyze and visualize the information acquired. The most commonly used image processing methods to analyze UAV imagery for weed mapping are photogrammetry and machine learning. Photogrammetry regards the accurate reconstruction of a scene or an object from several overlapping pictures. Photogrammetric techniques are very commonly used in all types of applications as they are also required to create vegetation indices maps. However, photogrammetric techniques are in most cases used to compliment other types of data processing methods [29].

Machine learning is used to process the data acquired, for prediction and/or identification purposes, with great results in many domains. Machine learning techniques are often applied in precision agriculture to exploit the information from the large amount of data acquired by the UAVs. Machine learning is able to estimate some parameters regarding the crop growth rate, detect diseases, or even identify/ discriminate objects in the images [30].

**189**

**Figure 9.**

*weeds, and bare soil. Source: Peña [31].*

**Figure 8.**

*Variable Rate Application of Herbicides for Weed Management in Pre- and Postemergence*

those based on object-based image analysis (OBIA). Weed detection with UAVs based on object-based image analysis appears to be at an advanced stage and can be

In an example of weed mapping performed on corn, an UAV coupled with a six-band multispectral camera (visible and near infrared range) was used to map

After mapping, an OBIA procedure processes the data and generates a classifica-

The identification and delimitation of the weeds allows generating maps showing the infestation level (**Figure 10**). The information of this map can be integrated into VRA system and used for POST herbicide application. In this study, weed-free areas corresponded to 23% and areas with low infestation (<5% of weeds) to 47% of the total, indicating a high potential to reduce herbicide application [31].

The most promising technique for weed mapping is machine learning, especially

*DOI: http://dx.doi.org/10.5772/intechopen.93558*

used for specific weed management.

tion of weed, crop, and bare soil (**Figure 9**).

*Unmanned aerial vehicle (UAV) used for weed mapping. Source: Peña [31].*

*Partial view of the outputs of the object-based image analysis (OBIA) procedure: classified image with crop,* 

the area (**Figure 8**).

*Variable Rate Application of Herbicides for Weed Management in Pre- and Postemergence DOI: http://dx.doi.org/10.5772/intechopen.93558*

The most promising technique for weed mapping is machine learning, especially those based on object-based image analysis (OBIA). Weed detection with UAVs based on object-based image analysis appears to be at an advanced stage and can be used for specific weed management.

In an example of weed mapping performed on corn, an UAV coupled with a six-band multispectral camera (visible and near infrared range) was used to map the area (**Figure 8**).

After mapping, an OBIA procedure processes the data and generates a classification of weed, crop, and bare soil (**Figure 9**).

The identification and delimitation of the weeds allows generating maps showing the infestation level (**Figure 10**). The information of this map can be integrated into VRA system and used for POST herbicide application. In this study, weed-free areas corresponded to 23% and areas with low infestation (<5% of weeds) to 47% of the total, indicating a high potential to reduce herbicide application [31].

#### **Figure 8.**

*Pests, Weeds and Diseases in Agricultural Crop and Animal Husbandry Production*

different rates [28].

with the other types of sensors.

toring other characteristics of the crops [26–28].

other types of data processing methods [29].

discriminate objects in the images [30].

they are more efficient than the ground systems as they can cover a large field in a short amount of time and in a non-destructive way, which is very important. UAVs can gather images and derive data from the whole field that can be used to generate a precise weed cover map depicting the spots where the herbicide are needed in

A variety of different types of sensors can be used in an agricultural UAV depending on the different vegetation parameters that should be monitored. The main sensors used that meet the limitations mentioned above are: visible light sensors, red, green, and blue (RGB) color model, multispectral sensors, hyperspectral sensors, and thermal sensors. RGB are relatively low cost compared to the other types and can acquire high resolution images, are easy to use and operate, and are lightweight [29]. In addition, the information acquired requires simple processing. However, they are inadequate for analyzing a lot of vegetation parameters that require spectral information in the non-visible spectrum. Thus, commonly are used

Multispectral or hyperspectral imaging sensors can acquire information about

UAVs can acquire information for various features of the cultivated field by using specialized sensors. However, as mentioned above, there is still no standardized workflow or well-established techniques to analyze and visualize the information acquired. The most commonly used image processing methods to analyze UAV imagery for weed mapping are photogrammetry and machine learning. Photogrammetry regards the accurate reconstruction of a scene or an object from several overlapping pictures. Photogrammetric techniques are very commonly used in all types of applications as they are also required to create vegetation indices maps. However, photogrammetric techniques are in most cases used to compliment

Machine learning is used to process the data acquired, for prediction and/or identification purposes, with great results in many domains. Machine learning techniques are often applied in precision agriculture to exploit the information from the large amount of data acquired by the UAVs. Machine learning is able to estimate some parameters regarding the crop growth rate, detect diseases, or even identify/

the vegetation's spectral absorption and reflection on several bands. Spectral information can be significantly helpful in assessing a lot of biological and physical characteristics of the plants. This information is important to determinate which weed species are in the field [30]. Multispectral and hyperspectral sensors are frequently used, despite their higher costs. However, a drawback of these sensors arises from the fact that it is required to apply more complex preprocessing methods in order to extract useful information from the captured images. The preprocessing procedure of spectral images often contains the radiometric calibration, geometric correction, image fusion, and image enhancement. The main difference between multispectral and hyperspectral sensors is the number of bands (or channels) that each sensor can capture and the width of the bands. Multispectral sensors capture 5–12 channels, while hyperspectral images can usually capture hundreds or thousands of bands, but in a narrower bandwidth. Multispectral sensors are used much more frequently than hyperspectral sensors due to their lower cost, but hyperspectral technology appears to have a lot of potential and is considered the future trend for crop phenotyping research. Thermal infrared sensors capture information about the temperature of the objects and generate images displaying them based on this information and not their visible properties. This type of sensors is used for very specific applications (irrigation management). As a result, they are not frequently used in remotely piloted aircraft applications of UAV systems that focus on moni-

**188**

*Unmanned aerial vehicle (UAV) used for weed mapping. Source: Peña [31].*

#### **Figure 9.**

*Partial view of the outputs of the object-based image analysis (OBIA) procedure: classified image with crop, weeds, and bare soil. Source: Peña [31].*

#### **Figure 10.**

*Partial view of the outputs of the OBIA procedure: weed coverage map showing three levels of infestation (low, moderate, and high), crop rows, and weed-free zones. Source: Peña [31].*

When data collection and map generation is done for POST herbicide application, the whole process must be done as quickly as possible because in a few days, the weed dynamics can be changed and infestation levels can increase, making the recommendation map obsolete.

#### **4.2 Sensor based: real time**

When applying POST herbicides using a real-time based sensor method, there is no need of a prior area mapping. Spraying is based on sensors attached to the sprayer responsible for detecting weeds and applying the herbicide dose. In **Figure 11**, there is a basic model for this application type.

In real-time-based sensor method, the optical sensor collects data that are immediately processed by the computer, where the locations and doses to be applied are determined. This information is sent as a command to a nozzle controller.

#### **Figure 11.**

*Sensor-based VRA model for POST herbicide application. The system includes a multiple-camera vision system, a ground speed sensor, and nozzle controller. Source: Tian [32].*

**191**

**Figure 12.**

*Variable Rate Application of Herbicides for Weed Management in Pre- and Postemergence*

vegetatively, in view of their stability in spatial distribution [33].

the most widespread are Weed-it and WeedSeeker.

ing to the size of the plant (**Figure 13**) [34].

*4.2.1 WEED-IT*

In the spray boom, each nozzle is opened or closed by a solenoid valve connected to the controller, so that the nozzle controller can vary the flow applied or the total opening and closing of each nozzle. The presence of a GPS system is not essential for the operation of system, but it does provide guidance to the machinery operator and is useful for recording sprayed areas. The database can be used to improve weed control in the following years, especially for perennial species that reproduce

Depending on the model, the system components can vary in several characteristics. Optical sensors can be multispectral or infrared. The software can be composed from algorithms that can only identify green plants to deep neural networks that have the ability to learn to differentiate weed species. The controller can only open or close a spray nozzle or it can even coordinate the herbicide mixture and control the alternating flow of dozens of nozzles. The variations are huge, and the more research evolves, the greater the accuracy and reliability of the VRA [7, 32, 33]. Commercially, some companies have consolidated in recent years with VRA systems for application in POST with sensor methods based on real time. Among

WEED-IT is a high-performance localized spraying system, formed by chlorophyll detection sensors and extremely fast valves to guarantee application only where necessary (**Figure 12**). The system is based on the principle of chlorophyll fluorescence: a light source in the set of sensors emits a constant beam of infrared light that is absorbed by the plants chlorophyll and re-emits near infrared light (NIR). This emission is detected by the sensors by performing 40,000 readings per second and capture even the lowest chlorophyll fluorescence emissions activating the nozzle set only on the identified weeds, applying only what is necessary, accord-

The system can be installed in self-propelled and trailed sprayers, operating at speeds of up to 25 km h−1. In the spray bar, each sensor is responsible for covering 1 m in width and independently activating up to five nozzles with an opening time of 1 ms. Its valves have a system for modulating the width of the energy pulses that generate extremely rapid interruptions in the spray nozzle outlet; the greater the

In curves or maneuvers, the speed on the outside of the bar is greater than the inside; the system is able to correct the flow along the bar to apply equal amounts of

*WEED-IT performing application with weed detection by infrared sensors. Source: SmartSensing [34].*

number of interruptions, the lower the applied dose (**Figure 14**) [34].

herbicide even in curves or with speed variations (**Figure 15**).

*DOI: http://dx.doi.org/10.5772/intechopen.93558*

#### *Variable Rate Application of Herbicides for Weed Management in Pre- and Postemergence DOI: http://dx.doi.org/10.5772/intechopen.93558*

In the spray boom, each nozzle is opened or closed by a solenoid valve connected to the controller, so that the nozzle controller can vary the flow applied or the total opening and closing of each nozzle. The presence of a GPS system is not essential for the operation of system, but it does provide guidance to the machinery operator and is useful for recording sprayed areas. The database can be used to improve weed control in the following years, especially for perennial species that reproduce vegetatively, in view of their stability in spatial distribution [33].

Depending on the model, the system components can vary in several characteristics. Optical sensors can be multispectral or infrared. The software can be composed from algorithms that can only identify green plants to deep neural networks that have the ability to learn to differentiate weed species. The controller can only open or close a spray nozzle or it can even coordinate the herbicide mixture and control the alternating flow of dozens of nozzles. The variations are huge, and the more research evolves, the greater the accuracy and reliability of the VRA [7, 32, 33].

Commercially, some companies have consolidated in recent years with VRA systems for application in POST with sensor methods based on real time. Among the most widespread are Weed-it and WeedSeeker.

#### *4.2.1 WEED-IT*

*Pests, Weeds and Diseases in Agricultural Crop and Animal Husbandry Production*

When data collection and map generation is done for POST herbicide application, the whole process must be done as quickly as possible because in a few days, the weed dynamics can be changed and infestation levels can increase, making the

*Partial view of the outputs of the OBIA procedure: weed coverage map showing three levels of infestation (low,* 

When applying POST herbicides using a real-time based sensor method, there is no need of a prior area mapping. Spraying is based on sensors attached to the sprayer responsible for detecting weeds and applying the herbicide dose. In

In real-time-based sensor method, the optical sensor collects data that are immediately processed by the computer, where the locations and doses to be applied are determined. This information is sent as a command to a nozzle controller.

*Sensor-based VRA model for POST herbicide application. The system includes a multiple-camera vision* 

*system, a ground speed sensor, and nozzle controller. Source: Tian [32].*

**Figure 11**, there is a basic model for this application type.

*moderate, and high), crop rows, and weed-free zones. Source: Peña [31].*

recommendation map obsolete.

**4.2 Sensor based: real time**

**Figure 10.**

**190**

**Figure 11.**

WEED-IT is a high-performance localized spraying system, formed by chlorophyll detection sensors and extremely fast valves to guarantee application only where necessary (**Figure 12**). The system is based on the principle of chlorophyll fluorescence: a light source in the set of sensors emits a constant beam of infrared light that is absorbed by the plants chlorophyll and re-emits near infrared light (NIR). This emission is detected by the sensors by performing 40,000 readings per second and capture even the lowest chlorophyll fluorescence emissions activating the nozzle set only on the identified weeds, applying only what is necessary, according to the size of the plant (**Figure 13**) [34].

The system can be installed in self-propelled and trailed sprayers, operating at speeds of up to 25 km h−1. In the spray bar, each sensor is responsible for covering 1 m in width and independently activating up to five nozzles with an opening time of 1 ms. Its valves have a system for modulating the width of the energy pulses that generate extremely rapid interruptions in the spray nozzle outlet; the greater the number of interruptions, the lower the applied dose (**Figure 14**) [34].

In curves or maneuvers, the speed on the outside of the bar is greater than the inside; the system is able to correct the flow along the bar to apply equal amounts of herbicide even in curves or with speed variations (**Figure 15**).

#### *Pests, Weeds and Diseases in Agricultural Crop and Animal Husbandry Production*

**Figure 13.** *WEED-IT operating system. Source: SmartSensing [34].*

**Figure 14.** *WEED-IT valve system modulation. Source: SmartSensing [34].*

**Figure 15.** *Differential flow compensation system. Source: SmartSensing [34].*

The system has an important limitation. As the sensor is based only on the chlorophyll fluorescence, the system is not able to differentiate the crop and weeds, both are interpreted as living plants. Therefore, it is necessary to be careful with the application of nonselective herbicides in POST, as the crop will certainly be sprayed together with weeds.

### *4.2.2 WeedSeeker*

The WeedSeeker is another widely used commercial system that has the same WEED-IT operate principle, where a sensor emits red and near infrared light and a

**193**

**Figure 17.**

**Figure 16.**

*Variable Rate Application of Herbicides for Weed Management in Pre- and Postemergence*

photodiode detects the intensity of the reflected light (**Figure 16**). Afterward, the reading is converted into a command to apply or not the herbicide (**Figure 17**) [35]. The system can be operated at speeds of 20 km h−1 installed in trailed and self-propelled sprayers. Nozzles are opened by solenoid valves connected to a central controller. The sensor spacing is 38 cm, and each sensor controls one spray nozzle. Although WEED-IT and WeedSeeker have many similarities, some aspects differentiate the two systems. The WeedSeeker requires a prior calibration of the sensors in order for the system to operate correctly, while the WEED-IT does not require any calibration [35]. As both systems have own light source, they can perform applications at night. Both are highly efficient systems that fulfill your proposals well. There are few studies that compare two systems. In a study focused on methods of comparing commercial precision spraying technology, the authors compared the efficiency and precision of WEED-IT and WeedSeeker and however, this comparison was only undertaken with a 0.16 ha−1. In this way, WEED-IT can be more efficient for

*DOI: http://dx.doi.org/10.5772/intechopen.93558*

identifying newly emerged plants [35].

*How a WeedSeeker sensor works. Source: Trimble Agriculture [35].*

*WeedSeeker spray nozzles applying herbicide only to weeds. Source: Trimble Agriculture [35].*

#### *Variable Rate Application of Herbicides for Weed Management in Pre- and Postemergence DOI: http://dx.doi.org/10.5772/intechopen.93558*

photodiode detects the intensity of the reflected light (**Figure 16**). Afterward, the reading is converted into a command to apply or not the herbicide (**Figure 17**) [35].

The system can be operated at speeds of 20 km h−1 installed in trailed and self-propelled sprayers. Nozzles are opened by solenoid valves connected to a central controller. The sensor spacing is 38 cm, and each sensor controls one spray nozzle. Although WEED-IT and WeedSeeker have many similarities, some aspects differentiate the two systems. The WeedSeeker requires a prior calibration of the sensors in order for the system to operate correctly, while the WEED-IT does not require any calibration [35].

As both systems have own light source, they can perform applications at night. Both are highly efficient systems that fulfill your proposals well. There are few studies that compare two systems. In a study focused on methods of comparing commercial precision spraying technology, the authors compared the efficiency and precision of WEED-IT and WeedSeeker and however, this comparison was only undertaken with a 0.16 ha−1. In this way, WEED-IT can be more efficient for identifying newly emerged plants [35].

**Figure 16.**

*Pests, Weeds and Diseases in Agricultural Crop and Animal Husbandry Production*

The system has an important limitation. As the sensor is based only on the chlorophyll fluorescence, the system is not able to differentiate the crop and weeds, both are interpreted as living plants. Therefore, it is necessary to be careful with the application of nonselective herbicides in POST, as the crop will certainly be sprayed

The WeedSeeker is another widely used commercial system that has the same WEED-IT operate principle, where a sensor emits red and near infrared light and a

**192**

together with weeds.

*4.2.2 WeedSeeker*

**Figure 15.**

**Figure 14.**

**Figure 13.**

*WEED-IT valve system modulation. Source: SmartSensing [34].*

*WEED-IT operating system. Source: SmartSensing [34].*

*Differential flow compensation system. Source: SmartSensing [34].*

*How a WeedSeeker sensor works. Source: Trimble Agriculture [35].*

### **4.3 Robots for variable rate application (VRA) in postemergence (POST)**

Use of autonomous agricultural robots has an interesting potential as a valuable technological tool for precision agriculture, bringing the advantage of being able to make use of the various theories in robotic control, already grounded and consolidated for applications in several other areas [36]. The main characteristic that differentiates an agricultural robot from a simple machine or implement is the freedom degree and autonomy possessed by the robot, including the need for human operation. As agricultural robots must have a high degree of autonomy, tools are necessary so that they can distinguish targets and culture in the field, as well as to orient themselves spatially during movement. The way the distinction is made is through sensors. The main sensors used are GPS real-time kinematic (RTK), cameras, gyroscope, strobe, and proximity [36–38].

The recent trend in the development of mobile robots and autonomous vehicles to perform specific tasks is mainly guided by improving efficiency and leading to operating gains (reduces soil compaction, absence of operator) when compared to the use of large machines [39]. Although much smaller than conventional agricultural machines, they can act cooperatively and perform tasks such as spraying pesticides that pose risks to humans [40]. Sprayers coupled to robots can direct spray nozzles to weeds through a computer vision system. Some models use photovoltaic plates to take advantage of solar energy and reduce or eliminate fossil fuel consumption. With all the advantages related to the autonomy and efficiency of agricultural robots, the farmer can direct his time and efforts toward other agricultural activities such as negotiating sales contracts and making investment decisions.

Robots provide precision spraying, realizing the collection of weed position and incidence information in real time and transmitting them to an atomizer or sprayer that regulates the need for more or less herbicide. Despite having many advantages, the use of robots still has points to be improved, among them are the following:


The current limitations present in agricultural robots are being resolved with the evolution of the available technology, since the optimization of sensors and algorithms occurs constantly, while in a few years, these limitations can be overcome. Artificial intelligence used in agricultural robots is a way of recognizing patterns so that the computer can identify weeds, pests, disease symptoms, nutritional deficiency, degree of maturation, and cut-off point in the harvest, among others. In a simplified way, artificial intelligence consists of providing the machine with as many examples of situations and decisions as possible, whether historical or simulated based on existing knowledge, so that when faced with similar circumstances, it can make a decision [37, 38]. There are several examples of robots currently used in VRA, two of which are described below.

#### *4.3.1 Robot for Intelligent Perception and Precision Application (RIPPA)*

The Robot for Intelligent Perception and Precision Application (RIPPA) is an autonomous system developed by the University of Sydney for detecting weeds and applying herbicides in microdoses (**Figure 18**) [41].

**195**

**Table 3.**

*Source: Sukkarieh [42].*

*Specification description of RIPPA.*

*Variable Rate Application of Herbicides for Weed Management in Pre- and Postemergence*

The system has infrared and monochromatic sensors working with neural networks that make it possible to differentiate between crop and weed. In this way, the application and efficiency of the system are much more accurate. Due to its small size and high precision, the system is suitable for smaller areas, such as horticulture. The RIPPA is powered by solar energy through solar panels on the top of the machine. The system also has a sensor for collecting moisture and soil temperature, which makes data collection a little more complete, generating .XLS files so that the producer can create a database with information from his area. **Table 3** contains

With characteristics similar to RIPPA, BoniRob (**Figure 19**) was developed by the partnership between the companies BOSCH and AMAZONE, in Germany. It is slightly larger than RIPPA, but it is still smaller than a small car and is capable of applying localized pesticides, collecting soil samples, and analyzing to obtain real-

*DOI: http://dx.doi.org/10.5772/intechopen.93558*

some additional information from RIPPA [41, 42].

time characteristics such as pH and phosphorus levels [43].

**Specification description Value** Track width 1.52 m Max crop height 0.6 m (adjustable) IP rating IP65 Mass (no payload) Approx. 275 kg

Idle discharge time (no solar) 43 hours

Max payload 100 kg at max operating grade (12°) Charge-time from empty > 2 hours (dependent on charger)

Driving discharge time (0.5 m/s, no solar, no payload) 21.5 hours Max area traversed per charge (no solar) 8 hectares (~10 hours at 1.6 m/s)

*4.3.2 BoniRob*

**Figure 18.**

*RIPPA robot model. Source: Sukkarieh [42].*

*Variable Rate Application of Herbicides for Weed Management in Pre- and Postemergence DOI: http://dx.doi.org/10.5772/intechopen.93558*

#### **Figure 18.** *RIPPA robot model. Source: Sukkarieh [42].*

The system has infrared and monochromatic sensors working with neural networks that make it possible to differentiate between crop and weed. In this way, the application and efficiency of the system are much more accurate. Due to its small size and high precision, the system is suitable for smaller areas, such as horticulture. The RIPPA is powered by solar energy through solar panels on the top of the machine. The system also has a sensor for collecting moisture and soil temperature, which makes data collection a little more complete, generating .XLS files so that the producer can create a database with information from his area. **Table 3** contains some additional information from RIPPA [41, 42].

### *4.3.2 BoniRob*

*Pests, Weeds and Diseases in Agricultural Crop and Animal Husbandry Production*

cameras, gyroscope, strobe, and proximity [36–38].

a.Low autonomy compared to conventional machinery

b.Operational limitations in adverse field conditions

c.State of technological development

in VRA, two of which are described below.

applying herbicides in microdoses (**Figure 18**) [41].

*4.3.1 Robot for Intelligent Perception and Precision Application (RIPPA)*

The Robot for Intelligent Perception and Precision Application (RIPPA) is an autonomous system developed by the University of Sydney for detecting weeds and

**4.3 Robots for variable rate application (VRA) in postemergence (POST)**

Use of autonomous agricultural robots has an interesting potential as a valuable technological tool for precision agriculture, bringing the advantage of being able to make use of the various theories in robotic control, already grounded and consolidated for applications in several other areas [36]. The main characteristic that differentiates an agricultural robot from a simple machine or implement is the freedom degree and autonomy possessed by the robot, including the need for human operation. As agricultural robots must have a high degree of autonomy, tools are necessary so that they can distinguish targets and culture in the field, as well as to orient themselves spatially during movement. The way the distinction is made is through sensors. The main sensors used are GPS real-time kinematic (RTK),

The recent trend in the development of mobile robots and autonomous vehicles to perform specific tasks is mainly guided by improving efficiency and leading to operating gains (reduces soil compaction, absence of operator) when compared to the use of large machines [39]. Although much smaller than conventional agricultural machines, they can act cooperatively and perform tasks such as spraying pesticides that pose risks to humans [40]. Sprayers coupled to robots can direct spray nozzles to weeds through a computer vision system. Some models use photovoltaic plates to take advantage of solar energy and reduce or eliminate fossil fuel consumption. With all the advantages related to the autonomy and efficiency of agricultural robots, the farmer can direct his time and efforts toward other agricultural activities such as negotiating sales contracts and making investment decisions. Robots provide precision spraying, realizing the collection of weed position and incidence information in real time and transmitting them to an atomizer or sprayer that regulates the need for more or less herbicide. Despite having many advantages, the use of robots still has points to be improved, among them are the following:

The current limitations present in agricultural robots are being resolved with the evolution of the available technology, since the optimization of sensors and algorithms occurs constantly, while in a few years, these limitations can be overcome. Artificial intelligence used in agricultural robots is a way of recognizing patterns so that the computer can identify weeds, pests, disease symptoms, nutritional deficiency, degree of maturation, and cut-off point in the harvest, among others. In a simplified way, artificial intelligence consists of providing the machine with as many examples of situations and decisions as possible, whether historical or simulated based on existing knowledge, so that when faced with similar circumstances, it can make a decision [37, 38]. There are several examples of robots currently used

**194**

With characteristics similar to RIPPA, BoniRob (**Figure 19**) was developed by the partnership between the companies BOSCH and AMAZONE, in Germany. It is slightly larger than RIPPA, but it is still smaller than a small car and is capable of applying localized pesticides, collecting soil samples, and analyzing to obtain realtime characteristics such as pH and phosphorus levels [43].


#### **Table 3.**

*Specification description of RIPPA.*

**Figure 19.** *BoniRob model. Source: Sellmann et al. [44].*

To ensure its operation, BoniRob has a set of cameras and sensors (**Figure 20**) that work as follows: camera "a" points to the top of the plant with the function of detecting and locating it; camera "b" is positioned to obtain a side view of the plants looking for overlapping plants. In "c," we have a set of light-emitting diodes (LEDs) that are responsible for emitting red and infrared light to assist the cameras when capturing photos. There is also a third camera, which has a high frame rate and resolution (higher than cameras "a" and "b") attached to the sensor responsible for spraying. This sensor, to maintain accuracy in capturing images and also during the application of pesticides, has a strobe that allows, even with variations in the terrain, the camera and the spray tip to remain in the desired position [44].

When it comes to artificial intelligence, based on the culture and species of plants you want to work with and control, machine learning takes place through the developed algorithm and is trained based on obtaining images (millions of them) that allow you to characterize the plants according to their shape, size, and color, among other parameters, allowing them to be recognized and distinguished in the face of a possible action such as spraying it or not [44].

As mentioned earlier, the versatility of agricultural robots is essential, since in the field, the conditions are highly heterogeneous. For this reason, many of these machines allow the installation of modules that perform different functions. In the case of BoniRob, we have a module for phenotypic recognition, a penetrometer, and a localized spraying mode already developed, but there are numerous other possibilities for adaptation and creation based on the particular characteristics to which the use of the machine is intended [44]. Other models of agricultural robots are being developed and gradually made available on the market. A good example is Ecorobotix (**Figure 21**), which applies microdoses of herbicide and

**197**

below.

**Figure 21.**

*Variable Rate Application of Herbicides for Weed Management in Pre- and Postemergence*

works completely autonomously. Its use is recommended after an initial standard application of herbicide, in order to replace subsequent applications and thus save

*Ecorobotix components. (1) Photovoltaic panels, (2) camera and artificial vision for steering and detection, (3) navigation by GPS and sensors, (4) electrical drive system, (5) rapid robotic arms with sprayers, and (6) tanks* 

The market robots for herbicide application are still at the beginning of its development and consolidation, but it represents a new way of interacting with

In order for the VRA to happen efficiently, it is necessary to have a high control in the spraying system responsible for the application of the herbicide. Controllers can act by modifying the pressure at the spray nozzles, or they can change the herbicide concentrations and the water flow in real time. Some of these systems are more complex, while others are simpler. The main controllers will be discussed

In flow-based control system, only the flow and pressure are changed. There is no manipulation of the herbicide concentrations. The system has a flow meter, a speed sensor on the ground, and a servovalve with an electronic controller to apply the desired rate of the tank mixture (**Figure 22**). A microprocessor uses information about the width of the sprayer and the recommendation of the spray volume per hectare to calculate the flow rate appropriate for the current speed of the soil. The servovalve is opened or closed until equal amounts of herbicides are applied regardless of the speed of the machinery. If the controller can be integrated with a recommendation map system, a VRA can be done. These systems have the advantage of being reasonably simple. They are also able to make rate changes across the

Depending on the speed, problems with drift can occur, as the flow sensor and servovalve control the flow of the tank mixture, allowing variable pressure rates to be delivered to the spray nozzles. Thus, high speeds can represent an increase in the

pressure of the nozzles and a consequent decrease in the droplet spectrum.

agriculture, revolutionizing the relationship between man and the field.

*DOI: http://dx.doi.org/10.5772/intechopen.93558*

an important amount of herbicide [45].

*for two different products. Source: Ecorobotix [45].*

**5. Variable rate controllers**

**5.1 Flow-based control systems**

bar in 3 to 5 seconds [7, 46].

**Figure 20.** *BoniRob components. Source: Sellmann et al. [44].*

*Variable Rate Application of Herbicides for Weed Management in Pre- and Postemergence DOI: http://dx.doi.org/10.5772/intechopen.93558*

#### **Figure 21.**

*Pests, Weeds and Diseases in Agricultural Crop and Animal Husbandry Production*

To ensure its operation, BoniRob has a set of cameras and sensors (**Figure 20**) that work as follows: camera "a" points to the top of the plant with the function of detecting and locating it; camera "b" is positioned to obtain a side view of the plants looking for overlapping plants. In "c," we have a set of light-emitting diodes (LEDs) that are responsible for emitting red and infrared light to assist the cameras when capturing photos. There is also a third camera, which has a high frame rate and resolution (higher than cameras "a" and "b") attached to the sensor responsible for spraying. This sensor, to maintain accuracy in capturing images and also during the application of pesticides, has a strobe that allows, even with variations in the terrain, the camera and the spray tip to remain in the desired position [44]. When it comes to artificial intelligence, based on the culture and species of plants you want to work with and control, machine learning takes place through the developed algorithm and is trained based on obtaining images (millions of them) that allow you to characterize the plants according to their shape, size, and color, among other parameters, allowing them to be recognized and distinguished in the

As mentioned earlier, the versatility of agricultural robots is essential, since in the field, the conditions are highly heterogeneous. For this reason, many of these machines allow the installation of modules that perform different functions. In the case of BoniRob, we have a module for phenotypic recognition, a penetrometer, and a localized spraying mode already developed, but there are numerous other possibilities for adaptation and creation based on the particular characteristics to which the use of the machine is intended [44]. Other models of agricultural robots are being developed and gradually made available on the market. A good example is Ecorobotix (**Figure 21**), which applies microdoses of herbicide and

face of a possible action such as spraying it or not [44].

**196**

**Figure 20.**

*BoniRob components. Source: Sellmann et al. [44].*

**Figure 19.**

*BoniRob model. Source: Sellmann et al. [44].*

*Ecorobotix components. (1) Photovoltaic panels, (2) camera and artificial vision for steering and detection, (3) navigation by GPS and sensors, (4) electrical drive system, (5) rapid robotic arms with sprayers, and (6) tanks for two different products. Source: Ecorobotix [45].*

works completely autonomously. Its use is recommended after an initial standard application of herbicide, in order to replace subsequent applications and thus save an important amount of herbicide [45].

The market robots for herbicide application are still at the beginning of its development and consolidation, but it represents a new way of interacting with agriculture, revolutionizing the relationship between man and the field.

## **5. Variable rate controllers**

In order for the VRA to happen efficiently, it is necessary to have a high control in the spraying system responsible for the application of the herbicide. Controllers can act by modifying the pressure at the spray nozzles, or they can change the herbicide concentrations and the water flow in real time. Some of these systems are more complex, while others are simpler. The main controllers will be discussed below.

#### **5.1 Flow-based control systems**

In flow-based control system, only the flow and pressure are changed. There is no manipulation of the herbicide concentrations. The system has a flow meter, a speed sensor on the ground, and a servovalve with an electronic controller to apply the desired rate of the tank mixture (**Figure 22**). A microprocessor uses information about the width of the sprayer and the recommendation of the spray volume per hectare to calculate the flow rate appropriate for the current speed of the soil. The servovalve is opened or closed until equal amounts of herbicides are applied regardless of the speed of the machinery. If the controller can be integrated with a recommendation map system, a VRA can be done. These systems have the advantage of being reasonably simple. They are also able to make rate changes across the bar in 3 to 5 seconds [7, 46].

Depending on the speed, problems with drift can occur, as the flow sensor and servovalve control the flow of the tank mixture, allowing variable pressure rates to be delivered to the spray nozzles. Thus, high speeds can represent an increase in the pressure of the nozzles and a consequent decrease in the droplet spectrum.

#### **Figure 22.**

*VRA spraying system that is a flow-based system of application rate. Source: Grisso et al. [7].*

### **5.2 Chemical direct injection systems**

In this system, the mixture is prepared with direct injection of the chemical in a flow of water. This system (**Figure 23**) uses a controller and a pump to manage the chemical injection rate instead of the flow rate of a tank mix [46]. The water flow rate is constant and the herbicide injection rate is varied to accommodate changes in soil speed or changes in the prescribed rate.

With the chemical injection, there is no leftover mixture and the direct contact of the operator with toxic products is reduced [10]. The system allows you to control the desired size and spectrum of droplets, since the variation of the application rate does not depend on the flow and pressure on the spray nozzles. Its main disadvantage is the long transport delay between the chemical injection pump and the discharge nozzles at the ends of the boom.

**199**

*Variable Rate Application of Herbicides for Weed Management in Pre- and Postemergence*

In this system, there is control of the herbicide injection rate and water flow rate to respond to changes in speed or application rate. A control circuit manages the injection pump, while a second controller operates a servovalve to provide a corresponding water flow (**Figure 24**). Such a system provides a mixture of constant concentration. The system can have many of the advantages of the previous two systems. There is no leftover mixing; the operator is not exposed to chemicals in the tank mixing process; the variation from one rate to another occurs quickly. The disadvantages include related to the complex system, higher initial costs, problem in delivering variable rates of liquid through in the nozzle spray, and modulated

*A direct chemical injection system with carrier control. Source: Rashidi and Mohammadzamani [46].*

The variable rate application (VRA) of herbicides has great potential for use in agriculture because it allows better control of weeds at lower costs and reduction in the use of inputs and environmental contamination. The main techniques available are based on the generation of application maps and the use of sensors in real time to identify weed infestations, which can be used in the preemergence (PRE) and postemergence (POST) of weeds. Both modalities are equally important in integrated weed management. VRA systems still require relatively high investment, restricting their use. The constant improvement of the VRA should further increase its benefits and reduce the costs of adopting the system, allowing its use by more farmers. The use of precision agriculture in farming systems is a path of no return, in view of the conjuncture of food production needs and scarcity of natural resources. Thus, VRA tends to be used more and more frequently until possible complete replacement of the conventional way of using herbicides in agriculture.

*DOI: http://dx.doi.org/10.5772/intechopen.93558*

**5.3 Direct chemical injection with carrier control**

spraying nozzle control systems [10, 46].

**6. Conclusions**

**Figure 24.**

**Figure 23.**

*VRA spraying system that incorporates chemical injection technology. Source: Grisso et al. [7].*

*Variable Rate Application of Herbicides for Weed Management in Pre- and Postemergence DOI: http://dx.doi.org/10.5772/intechopen.93558*

**Figure 24.** *A direct chemical injection system with carrier control. Source: Rashidi and Mohammadzamani [46].*
