The previous sections of this chapter presented the trends and drivers of motorized agricultural mechanization, and discussed the role of digital technologies in transforming agriculture in terms of their potential to enhance precision agriculture and expand inclusive access to agricultural machinery. This section looks more closely at the current state of digital automation technologies in agriculture and the main drivers of adoption, based on available evidence.
Continued use of a technology is the best indicator that it has been beneficial for at least some agricultural producers and businesses.48 The literature on the evolution of digital automation in agriculture provides insights into its benefits, challenges and adoption trends. In summary, the adoption of digital automation technology in agriculture has been driven by two main forces: rising food demand in the face of decreasing natural resources; and developments in other sectors of the economy, which drive innovation in the agriculture sector.48
To understand trends in digital automation technologies in agriculture, information must be collated from a variety of sources, because data are sparse (especially in low- and middle-income countries) and, in addition, no country or organization systematically collects data on their use. An individual analysis is of limited value because of the specificity of the technology and the country involved. It is only when information is considered as a whole that patterns emerge. Table 2 presents selected milestones in digital automation in agriculture, listing the first mover of each technology. Dating the introduction of each technology at the producer level is not simple, therefore the dates, countries and technologies in the table are only indicative of general adoption patterns; indeed, no technology emerges fully developed from the laboratory or design studio before moving to the farm. On the contrary, technology adoption is an iterative process; it starts with basic research to show the potential application and then converts scientific ideas into usable commercial products. Building on Figure 2, Figure 5 provides further examples of technologies covered in this chapter, organized by agricultural production system. These do not mirror, but rather complement, the technologies in Table 2.
FIGURE 5 Selected digital technologies and robotics with artificial intelligence by agricultural production system
TABLE 2SELECTED MILESTONES IN DIGITAL AUTOMATION IN AGRICULTURE
SOURCE: Lowenberg-DeBoer, 2022.48
Advances in automation for livestock production
As illustrated in Table 2, some of the first digital automation technologies emerged in the livestock sector. Precision livestock farming is made possible by attaching sensors to animals or to barn equipment to operate climate control and monitor animals’ health status, movement and needs, including related to breeding.67 Several precision livestock technologies have been developed that facilitate management of individual animals based on electronic identification (EID) tagging, the most common being milking robots, which allow cows to be milked without direct human involvement. The conventional milking machine uses a vacuum technology but still requires a human operator to place it on and remove it from the animal. On the other hand, EID automates the process by allowing a milking robot to access a database of udder coordinates for specific cows.68 This fully automated system adapted to animal production has great prospects in terms of cost savings and raising productivity.69 However, the evidence of the monetary benefits of milking robots is mixed: some studies indicated a positive impact,70, 71, 72 while others found no financial gains compared with conventional milking machine systems.71 Therefore, it would seem that adoption is driven not only by monetary but also by social considerations such as increased flexibility in work schedules and better quality of life – factors particularly relevant on small and medium-sized farms. More recently, however, larger dairy farms (with over 1 000 cows) have joined medium-sized farms in adopting robotic milking systems due to labour shortages. Therefore, the decision to use robotic milking may be based on quite different considerations on larger dairy farms.48 Box 5 presents examples of digital automation of livestock production in Africa, Europe and Latin America and the Caribbean.
Box 5Digital automation of livestock production: examples from Latin America, Africa and Europe
The company Cattler originated in Argentina in 2019, but has since expanded its business to other countries, including Paraguay and Uruguay and, more recently, Brazil and the United States of America. It offers an automated farm management system for beef cattle farms based on satellite information and provides feedback and insights for improving management. The company targets medium-sized rather than the largest farms. According to the company, a key driver of adoption has been the need to simplify operations and make a return on investment.
In Burkina Faso and Mali, and soon the Niger, with the support of the Netherlands Development Organization, GARBAL provides highly contextualized advisory information on livestock and crop production, as well as on fodder, milk and cereal markets. With a specific focus on women and youth, the digital solutions offered help small-scale producers and pastoralists affected by climate change in the Sahel make decisions concerning grazing lands, herd migration, the weather, and various farming practices. Solutions rely on satellite imagery, mobile phone SMS, unstructured supplementary service data (USSD), and a call centre with local operators speaking local languages. The use of mobile phones makes the solution very accessible. It was driven by, inter alia, public–private partnerships, subsidies, engagement with local farmer and pastoralist organizations, and the bridging of traditional and scientific knowledge. Key challenges have been the need for highly context-specific solutions, the security situation in some countries, significant capacity-building requirements, connectivity and network reception problems, and data quality issues.
Lely, a family-owned company in the Netherlands, offers robotics, as well as management software solutions for dairy farming, targeting medium- to large-scale producers with more than 100 cows, but not to date the largest farms. The principal technologies adopted are stationary milking robots, followed by manure robots and feeding robots. Grass harvesting robots optimize grass production, while upcoming products focus on reducing emissions. This is complemented by management software for all farm operations, including information on animal status welfare. The technology proposed can address issues of limited labour availability, emissions regulations and animal welfare. Key drivers of adoption are energy efficiency, chemical use reduction and labour shortages.
SOURCE: Ceccarelli et al., 2022.2
Global sales of automatic milking systems (AMS) have grown from USD 1.2 billion in 2016 to USD 1.6 billion in 2019, which points to a growing demand, albeit concentrated in high-income countries, with countries such as Germany, the Netherlands and the United Kingdom being early adopters.73, 74 Indeed, while no statistics on adoption are available across different regions and countries, the evidence indicates that adoption is confined to high-income countries, mostly in Northern Europe.75 Demand is driven by lack of rural labour, coupled with a generational shift. Table 2 shows that the first commercial AMS was used in the Netherlands in 1992; it has since spread to other countries.69 The absence of data on low- and middle-income countries suggests the technology is almost absent there.48, 76
In addition to milking machines, there are also technologies for the automated feeding of varying amounts of concentrates to cows based on their milk production.77 The same applies to poultry, where feeding systems are based on bird weight and egg count, and computerized control of ventilation is based on temperature and humidity.78 However, data and evidence regarding their adoption trends and drivers are even scarcer.
Advances in automation for crop production
Automation of crop production involves the use of many precision agriculture technologies, namely VRT, GNSS, robots, drones and AI. These may require the collection of spatial data, based on a geographic information system (GIS), using information from crop simulation models to identify the amount of inputs necessary to maximize yield and profit.67 Underlying these applications are sensors, including proximal sensing (e.g. measurement of nitrogen in the soil) and remote sensing (e.g. satellite imaging). Depending on connectivity, operators can share these data with stakeholders via smartphones and user-friendly apps that present the data in a simple manner.35
Adoption varies by agricultural commodity, capital cost, wage rate and other economic factors. In any case, adoption by small-scale agricultural producers is negligible; this is because there is almost no research on its adaption to small-scale agriculture and it is not easy to transfer the technology from mechanized to non-mechanized operations.
GNSS and VRT, matched with motorized machinery, are the most widely used in crop production to enable autosteer and on-the-go application of inputs. One of the main drivers of adoption of GNSS-based technologies is their capacity during application of inputs (e.g. fertilizer) to eliminate both accidental skipping and overlapping of plants, which translates into input savings. Other drivers include reduced operator fatigue, ability of family members to work longer hours, flexibility in hiring drivers (since they do not need to be highly skilled or experienced), and environmental benefits (as there are fewer overlapping applications), in addition to other advantages difficult to quantify and more akin to side-effects of adoption. The fact that the benefits of GNSS guidance are quickly appreciated (e.g. input savings from reduction in overlap are almost immediate) and visible to both farmer and neighbours (e.g. weed strips from herbicide skips are frowned upon in the farming community) also aided adoption.48
VRT technologies reduce input application and optimize crop yields, which also brings environmental benefits, especially if they reduce over-application. There is mixed evidence regarding the increased profitability of VRT fertilizers,79, 80 and this explains the modest adoption worldwide of map-based VRT fertilizer – and then mostly where profitability is consistent (e.g. nitrogen application to sugar beet).
In the most advanced automation category, autonomous crop robots entered commercial use only very recently. They appear mostly in high-income countries (e.g. France) for weeding organic vegetables and sugar beet.81 Hands Free Hectare – a project established in the United Kingdom in 2016 to develop and showcase agricultural automation – marked the first public demonstration of autonomous crop machines taking part in producing and harvesting a commercial crop.64 Since then, manufacturers have announced autonomous machines (see Table 2), and over 40 start-ups are currently developing them. Autonomous crop robots are associated with labour saving, improved timing of operations, more accurate input application and reduced soil compaction, especially with smaller swarm robots. A review of 18 cases found that autonomous crop robots used for harvesting, seeding and weeding were economically feasible in certain circumstances.82
In some countries, autonomous crop machines require on-site human supervision at all times, in which case the farmer may be better off using conventional equipment.83 One study found that remote supervision (e.g. from the farm office) is optimal only if the autonomous operation is relatively trouble-free.84 It emphasized the need for greater AI capacity to enable the autonomous machine to resolve more issues without human intervention. Similarly, speed restrictions for autonomous crop machines, as exist in the United States of America, can make them unprofitable.85
There are proposals to develop small, low-cost autonomous crop machines for small- and medium-scale farms as part of the solution to the lack of agricultural labour in low- and middle-income countries, with potential benefits especially for rural youth.86, 87, 88, 89 Unfortunately, there are no feasibility analyses for low- and middle-income countries. Nevertheless, the available literature indicates that the adoption of autonomous robots in these countries has the following potential benefits: (i) reduced human labour requirement, where labour is scarce; (ii) lower costs and reduced economies of scale, ensuring accessibility of technologies to smaller farms using conventional mechanization; and, (iii) ability to use technologies in irregularly shaped fields in a cost-effective manner, avoiding the reshaping of rural landscapes into large rectangular fields (where traditional mechanization is most efficient), a process that disrupts communities.
Drones are used for information gathering and to automate input application, similarly to map-based VRT. However, their use is often subject to strict regulations due to concerns about excessive input application, pesticide drift and aviation hazards.90, 91 For example, in the United Kingdom, drones are only allowed to apply herbicides in inaccessible locations under restricted conditions. Conversely, Switzerland allows more flexible input application by drones, which may encourage other European countries to do the same.83, 92 About 14 percent of agricultural retailers in the United States of America provided drone input application services in 2021, expected to increase to 29 percent by 2024.92 Drone input application is also quite common in some middle-income countries, such as Brazil and China.93
Some lesser known advances in automation: aquaculture, forestry, and controlled environment crop production
Digital automation is on the rise in the aquaculture sector in response to labour scarcity and high wages. There is wide adoption of innovations that automate feeding and monitoring, despite their high investment costs, as they minimize labour and other variable production costs and reduce the labour requirements to a few highly skilled operators.94 Box 6 showcases recent aquaculture innovations in India and Mexico.
Box 6New aquaculture technologies: examples from India and Mexico
Aquaculture has already demonstrated its crucial role in global food security and nutrition, constituting one of the world’s largest sources of animal protein, with production growing at 7.5 percent per year since 1970.95 Given the capacity of aquaculture for further growth, but also the enormity of the environmental challenges the sector faces as it intensifies production, new sustainable aquaculture development strategies are necessary. Such strategies need to harness technical developments in, for example, feed, genetic selection, biosecurity and disease control, and digital innovation. This can, in turn, enhance precision, improve decision-making, facilitate autonomous and continuous monitoring of fish, and reduce dependencies on manual labour, thus improving staff safety, fish health and welfare, while also increasing productivity, yield and environmental sustainability.96
Aquaconnect in India is a case in point. Although India is one of the world’s largest aquaculture producers, harvesting 7 million tonnes in 2018,95 the industry is characterized by a lack of transparency and inefficient value chains. Aquaconnect uses artificial intelligence and satellite sensing technologies to monitor the performance of aquaculture farms and provide shrimp and fish farmers (mostly small- to medium-scale) with advice to increase productivity. This solution is combined with an omnichannel platform that sells farm inputs at affordable prices. It also bridges the gap between farmers and financial institutions and improves market linkages. These solutions are currently assisting over 60 000 fish and shrimp farmers across India to increase productivity, enhance market linkages and improve access to formal credit and insurance.1 In parallel, the Government of India has allocated about USD 3 billion for the modernization of agriculture, including value chains of aquaculture and fisheries, and expressed interest in supporting initiatives (e.g. start-ups) that implement technologies and promote innovation.
Another ambitious project that promises to transform the aquaculture industry is Shrimpbox, the world’s first robotic shrimp farm, developed in Oaxaca, Mexico (see Atarraya case study in Annex 1). The technology provides automated systems that can be monitored remotely with software capable of learning and making decisions. The systems are integrated with biocontrol based on microbial methods to reduce nitrate build-up, prevent diseases, and save water in shrimp production, leading to significant reductions in water consumption, labour requirements, risk of diseases, and losses.2 According to the creators of the technology, a robotic farm can produce as much in 0.5 ha as a traditional 100-ha farm, while using only 5 percent of the water and remaining antibiotic-free.97 Shrimpbox can farm shrimp in colder climates and without ocean access. This in turn means that fresh, high-quality shrimp can be delivered to regions that today depend on imports of frozen produce.
Aquaconnect and Shrimpbox are just two examples of new technologies set to make aquaculture a more sustainable, inclusive and efficient process. However, the priority should be to further develop aquaculture in Africa and in other regions where technological development is lagging and food insecurity and malnutrition are more severe.95
In forestry, many wood harvesting operations are already highly automated, using motorized machinery progressively upgraded with digital tools. More recently, mobile technologies, combined with virtual reality and remote sensing techniques, are paving the way for advanced automatic machines in the forest. Wood harvesters and forwarders – advanced machines used for log cutting and transport – are currently a major target of automation efforts.98 Novel, digital-based technologies are increasingly pervasive. A recent review revealed a strong emphasis on remote sensing-based innovations for forest monitoring, planning and management, where machine-learning techniques also play an important role in data collection, processing and analysis. The continued adoption of digital tools is likely to raise new questions about forest ecosystems as dynamic, social, ecological and technological landscapes. Future research should examine more closely how forestry researchers, managers and stakeholders can anticipate and adapt to both environmental and technological uncertainty in the forest ecosystem.99 Box 7 summarizes the evolution of the forestry sector in terms of mechanization and the potential for digital automation.
Box 7Evolution of the forestry sector: mechanization and digital automation
Historically, work in the forestry sector was physically hard and potentially dangerous, especially in the wood harvesting phase. Systems with low technological input required a special logging crew consisting of a logger and a supporting logger, with an additional group of workers to trim the branches. Once trimmed, another specialized team, comprising a marker, a cross-cutter and two to three draggers, would cross-cut the trees into logs.100 Because of the demanding labour requirements and the danger to workers, such manual logging methods are much less common now.
In the 1950s, a process began to upgrade the logging sector from reliance on mainly manual labour to mechanization and partial automation. Forestry harvesting can be divided into four distinct phases: felling of the trees, extraction from the forest, sorting and loading at a landing site, and transportation to the market. Harvester machines are now capable of multiple operations (felling, extracting, cross-cutting and sorting). Such machinery has resulted in significant increases in efficiency and improved working conditions. The advantages of mechanization and digital automation include the safety and comfort of the harvester operator. In the process, labour productivity has increased dramatically. In Sweden, productivity per worker increased sixfold from 1960 to 2010 (see figure).
FIGURE Standing volume of wood per working day in the Swedish forestry industry, rolling three-year average
Even in these more mechanized logging systems, labour typically represents about 30–40 percent of running costs in European countries.102 The work environment is stressful since operators need to make many decisions at a fast pace, manoeuvring complex machinery and identifying differences in log quality, thus limiting the number of hours they can work. Therefore, one way to increase productivity is to raise the level of automation. The adoption of autonomous equipment is driven by productivity and operational costs. Although an autonomous machine is generally slower than operator-handled equipment, it can still be more cost-effective; semi-autonomous machines may allow an operator to run multiple machines at the same time.
Most modern forestry machines can readily be converted for remote control at relatively low cost, with many working options already available. As noted, machine operation is typically slower – significantly so if the task is complex – and will not be adopted in forest operations based only on improved productivity. However, they could be considered for other reasons: to safeguard operator safety, or when a full-time on-site operator is underemployed.
There are currently no fully autonomous systems in timber harvesting. However, the extraction and subsequent transportation of stems and logs with GPS-guided systems have been identified as probable first robotic operations, soon achievable with modest research and development (R&D) investment. Plantation felling may also become economically feasible in the longer term, but this will require substantial R&D investment.103 Finally, road transport of harvested logs is an aspect of forest operations that needs improved productivity in the wood supply chain. There are rapid developments in driverless truck technology, with the benefit that autonomous trucking reduces labour requirements and hence costs. For truck movement off public highways, autonomous vehicles are already deployed in mining operations, making an expansion to forestry a real possibility.
New, more environmentally friendly harvesting systems are also under development. A walking harvester can now meet the challenge of harvesting on steep, sensitive or uneven forest terrain. One goal is to limit the negative impact on forest soils through spot-ground contact without leaving the continuous track of wheeled or tracked harvesters.103 While such systems are still far from the commercial stage, in New Zealand a swinging forest harvester functions while making no contact with forest soils. It operates independently of the terrain conditions (steepness, roughness, etc.) by staying above ground and moving from tree to tree using the trees themselves for support, thus reducing soil disturbance.104
These environmentally friendly developments can be valuable in forests where the use of motorized mechanization in harvesting can cause soil compaction and erosion, as well as biodiversity loss. Finally, if one considers that benefits provided by forests go far beyond wood production – they include carbon storage, non-wood forest products, erosion prevention, water purification and recreation – it is important to assess how, using sensors, digital automation can also increase the value of these benefits. One important example is the monitoring of deforestation, specifically illegal operations, using satellite data. The ability to monitor deforestation has greatly increased in terms of granularity of the data, which are now available globally at a 5-m resolution on a monthly basis. A concrete example in the Amazon Basin was the detection of forest loss due to oil palm plantations expanding into indigenous territory in Ecuador.105 Having such data freely available with global coverage is a great example of how digital solutions can be used to diagnose problems.
Another area where digital automation has potential is controlled environment agriculture (CEA), which includes greenhouse agriculture and vertical farming. Greenhouses are the most common form of CEA. By their very nature, they are amenable to environmental monitoring, control and optimization. Innovations in low-cost and low-power consumption sensors and instruments, communication devices, data processing and mobile applications, together with technological advances in design, simulation models and horticultural engineering, have led to a shift from conventional greenhouses to smart controlled environments.106 Start-ups specializing in CEA, such as Food Autonomy in Hungary, ioCrops in the Republic of Korea and UrbanaGrow in Chile, point to real potential in this area.2
Prior to undertaking large-scale commercial development, there is need for an accurate economic analysis, given the high start-up costs involved in automation of greenhouses and vertical farming.106 As with all the technologies presented in this chapter, the cost of increased automation relative to increased profitability is key and should be considered in future studies to justify greater levels of automation.
