APPLICATION OF DATA SCIENCE AND ARTIFICIAL INTELLIGENCE IN PRECISION LIVESTOCK FARMING (PLF)
Livestock farming is a critical component of the global food system, providing a significant portion of the world’s protein and other essential nutrients. However, traditional livestock farming methods are often wasteful, inefficient, and have negative environmental impacts. Precision livestock farming is a newer approach that seeks to address these challenges and improve the sustainability and productivity of livestock production. Around 1.3 billion people in developing countries depend on the livelihoods and food security provided by the livestock sector (FAO, 2017), which provides up to 50% of global agricultural gross domestic products (Herrero et al., 2016). Using concentrated feed, medications, and vaccinations, as well as enhancing infrastructure and feed efficiency, livestock farming has seen a major intensification in addition to a rise in the number of animals. Admittedly, the high density of animals raised in comparatively small areas leads to the deposition of significant amounts of excretory nitrogen, phosphorus, organic matter, and faecal microbes in the watershed, contaminating water systems globally through processes like surface water eutrophication and groundwater nitrate enrichment. The livestock industry uses a substantial amount of natural resources and has a big impact on the environment by changing the biogeochemical cycles of nitrogen, phosphorous, and carbon. This causes problems with air quality, global climate, soil quality, biodiversity, and water quality. As a result, livestock farming should be focused on developing more sustainable systems, using technologies and management techniques to do so. Since it allows for control over animal welfare and the farm’s microclimate, applying precision livestock farming could be a useful strategy for achieving environmental sustainability in livestock farming (Berckmans, 2014). PLF is defined as “the application of process engineering principles and techniques to livestock farming to automatically monitor, model, and manage animal production” and converting bio-response into relevant information that can be easily applied to different management aspects focusing both on the animal and on the environment (Tullo et al., 2017). By continuously monitoring health, welfare, production, reproduction, and environmental effect in real-time, PLF aims to manage individual animals. Continuous in this context indicates that PLF technology measures and analyses every second, around-the-clock, seven days a week. Early alerts that provide the farmer the chance to intervene as soon as the first signs of compromised welfare or health show are essential to the huge potential of PLF. Animal scientists, physiologists, veterinarians, ethologists, engineers, information and communication technology (ICT) specialists, and others work together in the multidisciplinary field of precision livestock farming as it uses technology and data to make informed decisions about livestock management. This includes the use of sensors, a global positioning system (GPS), and other tools to collect data on livestock behaviour, feed consumption, and other factors. This data is then analysed using artificial intelligence (AI) algorithms to make predictions and optimize production.
With the exploding human population, livestock sector is burdened both directly (increased demand for animal products) and indirectly (lack of fodder/feed resources/climate change) to improve the efficiency of production and minimize environmental degradation. Among the various interventions being explored to increase the productivity of livestock sector, the precision livestock farming (PLF) tools are increasingly being evolved as the answer for some of the key demand supply gap. The PLF builds upon application of mainly the Information and Communication Technology (ICT) to increase the efficiency of livestock production at the same time promoting better animal and human welfare. The PLF has bearings to revolutionize the livestock sector if main adoption issues are addressed.
Precision Livestock Farming (PLF)- Defined
It is still an arduous task to exactly define precision livestock farming, however, there exists a consensus that PLF refers to the application of technology to livestock management in such a way that animals could be constantly monitored in almost all aspects of production, reproduction and health, at the same time making livestock farming less human dependent. While some researchers suggest managing animals by ongoing, real-time monitoring of their health, welfare, reproduction, and influence on the environment, wherein “continuous monitoring” refers to logging and analysing of data every single second, at all hours of the day.
Other researchers, however consider PLF to be the use of technology by farmers to reduce their reliance on manual labour, assist them in (daily) management, and increase farm profitability.
Principles of Precision Livestock Farming (PLF)
The basic principle of PLF is simplification of data collection from animals and subsequent interpretation so that any identified problem could be solved in a time effective manner. With automation of animal monitoring, data is constantly logged for individual animals that allows for identification of behavioural changes, oestrus detection, early disease detection and disease forecast, among others. This allows the farmer/farm managers to precisely identify the issue and develop solutions accordingly.
How does a PLF system work?
It is important to recognise and categorise the animal bio-response (labelling), which includes animal behaviours, vocalisations, and metrics resulting from interactions between animals and their environment, to construct a PLF system. Then, an algorithm to simulate the target responses must be created. The method must then be evaluated to see if the predictions are accurate. By modelling, monitoring, and controlling animal bio-response in this way, it is feasible to give the farmer reliable feedback and improve his capacity to maintain contact with the animals.
According to the first European Conference PLF conference, held in 2003 in Berlin, a living organism is a system that is complex, individually unique, time-varying, and dynamic (CITD). A live organism is undoubtedly much more sophisticated than any mechanical, electronic, or ICT system. The kind of algorithms that must be created should take into account the CITD nature of living organisms. That indicates that algorithms employed to keep track of these constantly changing individuals must adjust to them over time or use techniques that can be applied in real-time in field applications. As a result, only a few techniques can be used to develop realtime monitoring solutions for both humans and animals.
Benefits of PLF
The primary goal of PLF is to make livestock farming more economically, socially and environmentally sustainable (Vranken and Berckmans, 2017).Feed is the most essential and expensive element in livestock production and it should be properly regulated to meet the energy needs of animals throughout their lifetimes while preventing overfeeding and nutrient waste in the environment. One of the key benefits of precision livestock farming is improved feed efficiency. Livestock farmers can use data from feed consumption sensors to determine the optimal feed formulation and feeding schedule for each animal. This can lead to significant reductions in feed waste and improved feed conversion ratios, which translate into cost savings and higher profits. Additionally, as food is provided individually, PF systems can decrease feed competition while also enhancing animal welfare. Precision livestock farming also has the potential to improve animal health and welfare. By monitoring the behaviour and movements of livestock, farmers can quickly identify signs of illness or injury and take action to prevent further harm. PLF tools have been extensively used to identify lameness, sub-acute rumen acidosis, reduced rumen functioning, and mastitis. This can lead to improved animal health and reduced veterinary costs. Additionally, improving animal health and lowering morbidity and mortality have the potential to reduce both methane and nitrous oxide from enteric fermentation and animal dung, so benefiting livestock producers. Unquestionably, lower mortality and morbidity lead to increased saleable production, which reduces non-carbon dioxide emissions per unit product. Another environmental advantage of preserving animal health is that it reduces the usage of antibiotics, which is an effective technique for limiting the antibiotic resistance phenomenon. Another benefit of precision livestock farming is its potential to reduce environmental impacts, e.g., implementation of a PLF system for ventilation control can cut ammonia emissions by 60-65% (Zhang et al., 2013). Rumination and feeding behaviour are connected to methane emissions and rely on the quality of the silage. Depending on the quality of the silage, it is possible to reduce methane emissions by up to 11% (Blaise et al., 2017; Warner et al., 2017). Additionally, precision livestock farming can help farmers use water and other resources more efficiently, reducing their environmental footprint. It has been predicted that if fertility is maintained at maximum levels, greenhouse gas emissions can be reduced by more than 20% per herd (Wilkinson and Garnsworthy, 2017). PLF technologies, according to various studies, can be critical in fertility herd management since the optimal time for insemination in cows and sheep can be recognised using devices, algorithms, and sensors, enhancing conception rates and hence the efficacy of mitigation efforts. Despite its potential benefits, precision livestock farming is still in the early stages of development. The cost of the technology and the need for specialised skills and training can present barriers to adoption for many farmers. However, as technology advances and costs decrease, precision livestock farming is likely to become more widespread and have a major impact on the livestock industry.
Areas of application
- Animal Monitoring – Over the last decade a lot of development have been made in the area of animal monitoring. These include the application of basic technology like RFID tags for animal identification, to advanced health monitors that enable farmer to identify even slight behavioural modifications at the level of individual animals. The use of RFID tags/microchips help identifying the animals without the need of constant handling/restraining, allows continuous monitoring and tracking of animals even in an extensive production set up. GPS tags are also being used in animal production to aid in animal tracking, which is particularly beneficial when it comes to large herds that are maintained on pasture. It also helps track the walking behaviour. Use of collars embedded with behaviour monitoring sensors can aid in identification of oestrus occurrences through measurement of rumination rate as well as feeding and resting behaviour of individual animals. Sensors like pedometers, accelerometers, and posture monitors are also being increasingly used in animal production which have application in almost all aspects of animal management.
- Animal Health and Welfare – Animal health has wide implications for the profitability and efficiency of livestock enterprises. With close association between animal and human health, the early forecast and efficient management of livestock diseases necessitates development of technologies that will aid in forewarning of effective states. In fact, this is of the key reasons for the development and further evolution of PLF technologies. As most diseases are easily treated on early detection, PLF technologies reduces losses incurred on account of diseases. Recent technologies make use of sensors, artificial intelligence, machine learning and big data analytics which help farmers address any health issues that become evident early on, in terms of altered behaviour, movement, feed and water intake and related deviations. With the constant data collection, and application of advanced analytics that allow prediction of deviations/aberrations, farmers are able to predict, identify and prevent possible outbreaks. These include application of thermal imaging to detect inflammation, lameness, and other irregularities. As thermal sensors allow continuous monitoring of temperature at almost all points of any animals, even slight day to day variations is easily detected and alerts the farmer of possible effective states, and has been found successful in detection of mastitis, lameness, hoof defects, among others. The other avenues include rumen sensors which are either telemetry-based pH monitoring system that helps track changes in rumen environment that can adversely impact feed digestion and lead to other health issues like ruminal acidosis/bloat etc, or temperature sensors that helps detecting alterations in rumen environment/heat stress or pressure sensors/motility sensors that helps in detection of bloat/aberrations in rumen motility etc and help disease forecast.
- Animal Growth and Nutrition – One of the more lucrative aspects of PLF technologies and its application to animal production is the precise livestock feeding which helps cater to individual animal’s feeding requirement matching it with nutrient supply, on the basis of real-time sensors. Almost all animal feeding operations are based on group feeding standards, as individually feeding and monitoring feed intake and growth is a tall task. PLF provides the way forward with individual feeding approach, as automatic monitoring of individual animals can be easily achieved in terms of feeding behaviour, feed intake, water intake, rumination time, among others. Numerous sensors have been developed over the years that helps measure varied aspects of nutrition like energy balance, feed digestion, feed degradation and energy expenditure, the data from these can be combined in multiple ways to optimize animal feeding in terms of nutrient intake, and feed efficiency. The other approach is application of photogrammetry which allows monitoring of accurate feed intake, feeding behaviour and aids in measurement of individual’s feed efficiency. However, the feeding interventions using ICT is still at incipient stage and more confined to research studies, so, there is a need to develop and integrate animal models with new and evolving technologies that are customized to deliver right amount of feed/nutrient of right feed to the right animal at the right time. PLF can ensure need based nutrient dissemination, thus, reducing wastage of feed and fodder resources owing to excessive supply to animals.
Advantages of Precision Livestock Farming
The PLT technologies have the potential to revolutionize livestock sector in terms of efficiency, profitability and sustainability at the same time accounting for welfare of both animal and humans alike. PLT technologies can help in:
- Optimized inputs and enhanced efficiency of livestock sector: PLF technologies ensures precise handling and management of various aspects of production, which enables proper utilization of resources, reduced wastage and consequently the cost of production declines. This translates into higher profitability which makes adoption of PLF technologies more lucrative to the farmers.
- Reduced labour requirement: With changing demographics, it is hard to find cost-efficient labour in primary sectors. PLF reduces the need for labour by encouraging greater automation and eliminating manual error in data collection, which helps in better decision making and efficient production.
- Enhanced animal health and welfare: PLF enables early disease forecast and helps in early detection of various ailments by detecting even slight alterations from normal behaviour/parameters of individual animals. This has widespread ramifications for animal and human health, farm economics and animal welfare. Furthermore, sensors monitoring behaviour are very helpful indicators of animal comfort and ultimately helps enhance animal welfare.
- Environmental implications: PLF technologies enable the monitoring of the amount of greenhouse gas (GHG) production like methane emissions from animal production systems, which helps incorporate changes that can mitigate GHG emissions. Indirectly, the PLF technologies reduces environmental impact of livestock sector by ensuring resource optimization and reducing resource wastage.
Disadvantages of precision livestock farming
- Affordability of advanced technologies, automation enabling devices like milking robots, relay feeders, etc is a key bottleneck when it comes to adoption.
- The technologies require a good share of customization when it comes to the varied nature of livestock enterprises in terms of species specificity/ specificity to physiological states.
- PLF technologies are still viewed in context of mere monitoring operations, if the collected data is not interpretated to drive gainful solution that betters animal production and welfare, the investment of capital in such technologies becomes a fallacy.
- The PLF technologies are hugely capital intensive and requires good infrastructural support.
- Data handling and interpretation is highly time-consuming process and requires dedicated workforce to identify problems based on analytics, develop efficient solutions and address the problems.
Challenges in adoption
With every new innovation, there is a fair amount of challenge associated with adoption. PLF is no different, it comes with its own share of challenges:
- Difficult to adopt in extensive systems: The adoption of PLF specially in developing nations become a challenging task because of the largely extensive nature of production systems, these systems make real time monitoring using wireless tech difficult owing to the limited range of most sensors. On the other hand, the microchips used in animal identification, wearable collars do provide a very good solution in extensive production but affordability becomes an issue. The whole market of software providers and technology innovators are monopolized by few big enterprises, and lack of market is a major deterrent that prevents entry of other players, making these solutions costlier.
- Lack of Standardization: Animals are considered as complex, individually different, time-varying, and dynamic (CITD) systems, which warrants a great amount of standardization if accurate results are to be derived. The whole spectrum of PLF technologies still lack standardization in terms of systems they are applied to, species they are adopted for, and physiological states of animals.
- Lack of skilled manpower: The conventional workforce employed in the livestock sector is not skilled to handle most technologies/collected data, especially the interpretation of collected data.
- High chances of equipment failure
- Intrusive nature of technologies: Most of the devices used for monitoring various activities are often very intrusive, that sometimes defeats the purpose of animal welfare by interfering with animal’s normal state.
Opportunities
The PLF technologies are still in their nascent phase when it comes to developing countries. These technologies are one of the key avenues that will transform how livestock keeping is done in the changing world, which puts significant pressure on livestock sector to feed an exploding world population at same time reducing environmental impacts. PLF will also generate new employment opportunities for a large workforce, from innovators to basic functionaries associated with livestock directly. However, to tap in the revolution, there is a lot of upgradations necessitated in terms of skilling and reskilling of labor force and evolve quality infrastructure that can address the needs of technology adaptation. With the rapid evolving nature of the sector, PLF can ensure food security while accounting for animal welfare.
Research on animal farming has significantly increased in recent years, with particular attention paid to sensor technology, data processing, transmission, and the use of AI models like machine learning (ML), deep learning (DL), and artificial neural networks (ANN). Identifying animals, identifying behaviours, tracking diseases, and maintaining environmental control are just a few of the difficulties faced by animal farming industry. By utilizing sensors, analyzing data, and leveraging AI models, researchers aim to improve animal farming practices. The application of AI technology in intensive animal farming has shown to be extremely advantageous since it enhances smart farming techniques, which eventually increase animal health, welfare, and economic outcome. Modern animal farming requires the use of cuttingedge technology like big data, AI, and ML. Big data offers scalability and accessibility by enabling the storing and administration of massive amounts of data on remote computers. At numerous points in the agricultural process, from data gathering to decision-making, AI technologies have been used. AI-enabled wearables, cameras, and sensors make it easier to monitor critical metrics, record animal behavioural patterns, and gather data automatically. Farmers may examine this real-time data using AI algorithms to spot anomalies, spot symptoms of disease or discomfort, and take immediate action. This proactive approach improves animal welfare and minimizes the economic impact of diseases.
Application of Big data and AI in precision livestock
Technologies for precision livestock husbandry are essential for supporting AI-based monitoring of animal welfare and health. These technologies include a wide range of topics, including sound analysis, tracking animal movement, eating behaviour, and water intake, and radio frequency identification (RFID) devices. By combining these technologies, AI systems can analyse the information gathered and offer detailed insights into the health and behaviour of animals.
AI and facial recognition technology
have shown great potential in various fields, including healthcare and animal farming. When applied to the early disease prediction in cattle, AI-based facial recognition systems can offer significant benefits. Early disease detection in cattle is crucial for prompt intervention and effective treatment, ultimately improving animal health and minimizing economic losses for farmers. Traditional methods of disease detection often rely on manual observations, which can be time-consuming and subjective. AI-based facial recognition systems provide a noninvasive and automated approach to monitor cattle health continuously. For instance, Identifying and separating sick cattle from the herd quickly can save thousands of dollars in medical treatment each year. Using MyAnIML’s technology, ranchers can identify, separate and treat cattle 2-3 days before symptoms arise. MyAnIML’s predictive health dataset leverages subtle changes in cow muzzles to detect cattle illnesses 2-3 days before they present symptoms. Therefore, AI and ML help cattle producers more effectively manage their Herd Health. The health and well-being of cow have a substantial influence on milk output. But maintaining the best possible cattle health often necessitates constant, round-the-clock observation of each animal, which drives up labour expenses by at least 30%. The world’s first dairy farmer’s assistant, Intelligent Dairy Assistant (Ida), was created by researchers to address this issue. Ida uses artificial intelligence (AI) to learn and understand cow behaviour and to give farmers useful operational insights. Ida turns field data collection into actionable information that aids farmers in making decisions on a daily basis by combining cutting-edge sensor technology, machine learning algorithms, and cloud computing. Ida goes above and beyond previous cow movement trackers that only provide mobility data to a central gathering location by performing intelligent pattern analysis. Ida can recognise and examine trends in cow behaviour, such as eating, resting, strolling, and drinking, by utilising AI. Then, any deviations or irregularities that would suggest possible health problems or decreased productivity are identified by comparing these behaviour patterns to established norms and algorithms. Ida equips farmers to act quickly by giving them real-time alerts and useful information, enhancing animal health and production.
Maintaining a vigilant watch over herd is crucial for ensuring optimal performance. Individual monitoring enables early heat cycle detection, illness diagnosis, and quick response for any other problems. However, it has become more difficult for farmers to set aside adequate time for attentive observation due to the growth in herd numbers and labour demands in recent decades. This problem is intended to be solved by U-motion®, a Cattle Behaviour Monitoring System powered by Artificial Intelligence (AI) (http://desamis.co.jp/en/). Cattle actions and behaviours, such as feeding, drinking, ruminating, moving, standing, and laying down, are continually recorded and observed by U-motion®. The application of AI technology enables this round-theclock surveillance. The system may detect trends, deviations, and abnormalities in the behaviours of particular cows or the herd as a whole by reviewing the collected data.
Enhancing artificial insemination (AI) success rates in dairy production is essential for maximising profitability. High conception rates are largely dependent on the timing of insemination, which works best when it takes place 8 to 12 hours following the onset of estrus. Accurately detecting cow estrus is one of the biggest hurdles in AI. Current estrus detection techniques are labor-intensive, not always accurate, and call for knowledgeable and experienced workers. Missed opportunities for insemination might result in financial losses for dairy producers if estrus is not detected in time or is misdiagnosed. The BOVINOSE is a significant initiative focused on developing an electronic nose designed to detect estrus in dairy cows. The primary goal is to develop a device that can precisely determine the best time for artificial insemination by identifying sex pheromones emitted by cows solely during estrus. The working prototype comprises of a probe, an array of sensors, and self-learning software for estrus prediction. The sensor array is in charge of detecting and analysing the chemical composition of these samples.
The accurate monitoring of feeding behavior, including rumination and eating patterns, is crucial for assessing the health, growth conditions, and early detection of diseases in cattle. Cattle behavioural data is gathered using a Noseband pressure sensor. The researchers developed a method that involved analysing local data variation, extracting frequency-domain features via Fast Fourier Transform (FFT), and then using the Extreme Gradient Boosting Algorithm (XGB) to categorise these extracted features into rumination and eating behaviours to address the challenge of accurately identifying feeding behaviour in cattle using data from a noseband pressure sensor. The ability to accurately identify individual animals is crucial for implementing precision management practices. In order to identify individual cattle using photos of their muzzles, 59 deep learning models’ performance and viability were evaluated. The findings showed promise for obtaining high accuracy and processing speed. As a way to improve precision livestock farming methods, automatic individual cow identification using video data has attracted a lot of interest. Researchers have created a unified deep learning architecture technique to accurately and effectively identify individual cattle from video data. The deep learning architecture combines multiple components, including Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM) networks, and self-attention mechanisms. These components work together to extract spatio-temporal features from the video data, enabling reliable identification of individual cattle. Forecasting the occurrence of disease based on meteorological and geographical data using ML approaches. A condition known as lumpy skin disease (LSD) is now growing quickly among cattle and water buffalo. It is an infectious, eruptive, and occasionally fatal illness characterized by skin nodules. It is feasible to obtain excellent accuracy in forecasting the presence of lumpy skin disease virus (LSDV) infection in unseen test data by using machine learning techniques and adding climatic and geographical factors as predictive variables. An example of such accomplishment is the use of an Artificial Neural Network (ANN) algorithm, which produced a 97% accuracy score.
Components required to develop AI-driven sensor in cattle
The creation of an AI-driven sensor system for cattle requires a number of elements. The following elements combine to gather data, process it with AI algorithms, and deliver useful information for managing cattle.
The following are the crucial elements:
1. Sensor Nodes: Sensor nodes are tiny electronic gadgets with a variety of sensors that may gather information about various behaviour and health characteristics of cattle. As examples of these sensors, we may mention gyroscopes, accelerometers, temperature sensors, humidity sensors, heart rate monitors, and GPS modules. In order to collect data in real-time, the sensor nodes are fastened to the cattle.
2. Data Acquisition System: A data acquisition system is responsible for collecting data from the sensor nodes. It typically includes hardware components such as microcontrollers or microprocessors that can communicate with the sensor nodes and gather the data they collect. The data acquisition system ensures reliable and continuous data transmission from the sensors.
3. Wireless Communication: Wireless communication is essential for transmitting the collected data from the sensor nodes to a central processing unit or cloud-based platform. This can be achieved using wireless technologies such as Wi-Fi, Bluetooth, Zigbee, or LoRaWAN. The choice of wireless communication depends on factors such as the range, power consumption, and data transfer rate required for the application
4. Data Storage: The obtained data must be saved in order to be processed and analysed further. This can be accomplished by local storage within the sensor nodes or through data transmission to a cloud-based storage system. Cloud storage provides scalability and accessibility, making it possible to easily retrieve and analyse data from various sensor nodes.
5. Artificial Intelligence Algorithms: AI algorithms are critical for analysing data and generating relevant insights. Deep learning and other machine learning (ML) methods aid in tasks such as activity recognition, behaviour detection, disease diagnosis, and prediction. Based on the patterns they identify, these algorithms can offer predictions or classifications.
6. Data Processing and Analytics: To extract usable information, the acquired data is processed and analysed using AI algorithms. This includes pre-processing raw data, extracting features, and applying AI models to tasks like as anomaly detection, health monitoring, and behaviour prediction. Data analytics approaches are used to gather insights and make educated livestock management decisions.
7. User Interface and Visualisation: The AI system should give farmers or veterinarians with a user-friendly interface for accessing and interpreting the acquired data and insights. Web-based dashboards, mobile applications, and other graphical user interfaces that provide information in a clear and actionable manner are examples of this.
In conclusion, precision livestock farming is a promising approach to improving the efficiency, sustainability, and productivity of livestock production. By using technology and data to make informed decisions, precision livestock farming has the potential to revolutionize the way we approach livestock management and ensure a more sustainable and profitable future for the livestock industry.
Compiled & Shared by- Team, LITD (Livestock Institute of Training & Development)
Image-Courtesy-Google
Reference-On Request.
Artificial Intelligence(AI): Future of Livestock farming in India
