Santiago A. Utsumi, PhD
Innovations for Sustainable
Farming & Ranching
Inspired by Nature, Driven by Science
New Frontiers for Digital Dairy Farming
As the era for digital farming unfolds, the adoption of robots for dairy farms reaches a new turning point. To put it in numbers, over 3.5 million cows will milk today by a robot, and anywhere between 1x to 5x. A story tale that was hard to believe only few decades ago. According to a recent market and trade report the business for robots closed at $ 6.5 billion for 2018, and is expected to reach 65% of the global dairy herd (today at 21 million cows) before this decade ends. Most of the demand is expected to come from a greater penetration of robots in large farms.
Robots introduce the automation of milking in the context of a fully voluntary process. This means that each cow in your herd will have the absolute freedom to milk by a robot at her desired pace (left). Freedom and flexibility to schedule chores usually rank high among those that switched from conventional milking to robotics. Other benefits of the technology may include (but it is not limited to) quarter milking (left), high process control, highly standardized milking routines, reduced farm labor, remote inspection for milk quality, cow health and breeding checks, and increased production through individualized flexible schedules for milking and concentrate feeding.
Indeed, robots are true Smart-Management Platforms that are opening new doors for Big Data-driven Digital Farming. Relevant to herd management is the fact that most robotic milking apps and software can capture, synthesize and transform (both directly and indirectly) very complex and large biological data sets into key performance indicators (KPIs) that staff and managers can check regularly to traction decision making, thereby making dairy management more predictable and reliable (Left),
Here I share what I see as the four golden rules I have learnt from several years of robotic milking research. Whether used for high cows on drylot yards, compost-bedded pack barns, freestall barns, or implemented for large herds of grazing cows, robots will do their job 24/7 as long as there is proper management. Be aware of managing cow traffic! (Advise #1). Traffic refers to the movement of cows with regard to robots, and as our research at MSU says the main drive for cow traffic is not just appetite but the rewarding desire for nutritious feed that truly nourishes and satiates. Remember that any feed manipulation made in the attempt to change cow traffic will do its job, but only within the set of those decision constraints and boundaries for given environmental factors, such as the type of traffic system used (free or guided), the barn design, your pasture and cow alleyways layout, or climatic factors.
Further, our observations at the MSU farm also show that any challenge imposed through feeding management, either due to challenging layouts (i.e. long walking distances to fresh grass, many turning angles over raceways, poor raceway grounds, etc), irregular and inconsistent feeding frequencies (i.e. delayed feeding intervals), or use of unpleasant feeds of reduced quality and preference (i.e. aged feed, bitter feed, moldy feed, etc.), will reduce cow traffic and the efficiency of robots. Thus, providing the needed cow comfort and luring cows to milk voluntarily, at ease pace, and according to unique needs for each cow, either nutritionally, metabolically, or even socially, is in fact the art and science for robotic milking (Advise #2).
Good robot management is that one that optimizes your goals for production and profit (i.e. milk revenue by the robot) while managing for cow comfort, health, and well being. This must be set through a round of compromises between stocking rates and individual production, and as your dairy cows transition from high and mid diets to low diets. Work and plan proactively with your team of nutritionists, agronomists, staff, and your robot supplier to come up with SOPs and transition plans that truly fit your own needs, experience, and goals (Advise #3).
Above I share the main synthesis for robotic research on cow traffic and feeding, adapted from previous modelling by Dr. Emilio Laca and Dr. Fred Provenza. The model workflow considers both cognitive (learning and memory) mechanisms and behavioral processes (animals' decision making) linking physical futures and attributes of a farm with effects of feeding management. Good provision and management facilitation of cow comfort (environment), together with traffic-based management of the cows' nutrition and feeding (internal state), and properly designed robot layouts (either pasture or barn) are the key for success (Advise #4).
Feeding the robotic dairy cow
Robots are gaining traction among those that constantly seek on innovations to balance for production, profit, flexible labor, better lifestyle and cows' well being, of course. Feeding in robotics doesn't appear to change the game for the nutrition of close ups, freshening, peak milk, mid lactation and dryoffs, but it does bring new dimensionalities regarding feeding management. Whether on high diets or on grass-based diets robotics introduces the need to incentivize cow traffic while managing to meet the uniqueness for cows and their specific requirements, nutritionally, metabolically and behaviorally. Likewise, robotics bases many tasks on digital programming and deep machine learning for sake of optimization, consistent standardization of milking routines, and to match deferentially milk and robot feed tables, either by grouping virtually or physical according to one or more functional parameter (parity, day in milk, milk yield, body condition, etc), When planned and executed consistently and according to predefined objectives and goals, tasks and KPIs for milk and feed tables and feed efficiency, must rank at the top of your dairy agenda Another interesting characteristic is the opportunity to differently feed cows, which is an improvement that goes well above and beyond the types of dairy cows used (high vs. low yield), the dairy systems implemented (pasture vs. conventional) or the chosen cow traffic design (free vs. guided robot layout).
How to feed?
Consider the following data for insights about feeding robotic cows, Here I share a comparison across diets and 'dairy feeding systems' from a farmlet study I 've conducted in 2012 with my colleague and MSU's Dairy Nutritionist Dr. Dave Beede. We compared robotics for a pasture-based feeding system (A-B system; 12 h breaks) with variable rate feeding (6 kg of milk per 1 kg of pellet feed) of a robot concentrate supplement (Figure a; 27 kg milk), or the same prescription table for concentrates and pasture-base (A-B system, 12 h) plus feeding of a partial mixed ration (PMR, maize silage based) in the evening (Figure b, 32 kg milk), or, the same prescription for the robot concentrate plus the same PMR, fed twice daily at 5% orts (Figure c; 41 kg milk). Multivariate analysis using structural equation modeling of 'big data sets' of daily cow records were used for prediction of standardized regression coefficients of variables, thereby coefficients closer to 1 will meant strong effects, whereas positive and negative coefficient symbols indicated differences in the direction for effects, respectively.
I would like to use results to draw on the following principles and recommendations. The first relates to the realization of complex and multivariate feeding systems, because not just one but many cow-, environment- and management- related variables (i.e. GxExM) will define the 'robotic success'. Consider 'Figure a' for a sec. As a cow's days in milk (DIM) increases and the cow moves into a longer lactation, the allocation for the robot supplement the cow receives decreases, thereby reducing over time the enticement for visiting a robot (defined here as AMS; Automatic Milking System). Likewise, increasing robot visits will mean greater milking frequency, but because feed incentives decline over time, cows on extended lactations will reduce significantly the number of milkings. The second concept is the need to match milk and feed tables, consistently and according to expectations for milk production. By going back to Figure a, notice that by increasing the allocation of robot supplements we can increase milkings, and by increasing milkings we can reach our objective of increasing milk yield. However, also notice that a more effective pathway to increase yield is through a direct increase of the robot concentrate supplement (mainly energy), as one would expect; yet also notice that both, the effect of supplements and milkings on yield differ markedly by feeding and diet context. Increasing the robot supplement will have a greater effect on milk for extensive (low energy) feeding systems, such as a grass-based system (Figure a), whereas the increase of milkings will have a much greater effect on intensified feeding systems using conventional PMR (Figure c). The hybrid grass-based plus PMR feeding system ranks (mechanistically) between the two systems at the extremes (Figure b). Finally, another dimentionality of robotic feeding, which is quite often confounded with the use of either guided or free cow traffic, is the ability to use 'variable rate' feeding as opposed to using a 'flat rate' feeding of robot concentrates, thereby allowing for increased control on robot feed conversion efficiency and for achievement of differential traffic and milking across cows..
How much to feed?
Dairy robotics is synonym of partial feeding systems. This means that the bulk of diets come either from the bunk or from grazed pasture, and the remainder from the robot. In some cases a different source for partial feeding of concentrates could be used through Smart-feeding stations, outside robots. So, how much to feed usually relates to your targeted objectives for cow traffic and nutrition, and this component must surpass decisions regarding the robot layout, whether your cows are being managed on grass, freestalls, compost barns, drylot yards, or whether cows are on a combination of any housing and pasture system. As a basic rule, any excess of partial feeding either from the bunk or through over-allocation of grass will trade off the value for robot feed, thereby reducing cow traffic and increasing the list of 'late' cows that must be brought for milking (left). Usually, these are your extended lactation cows, low yielding cows, attention or special needs cows (i.e. lame cows, cows in heat, or first lactation freshening cows on training), and cows with restricted robot feed, and usually when a flat low rate robot feed is used across all cows. Conversely, by changing the formulation of energy, protein, or energy/protein ratios on alternative feeds, either at the bunk (i.e. grass to concentrate ratios of PMR) or through accurately allocated grass (spatially and temporally), afferent signaling both for appetite and satiety can be targeted synchronously to better address cow traffic needs, thereby increasing cow movements while reducing the frequency and size of the 'late' cow list (right).
Where to feed?
This question is increasingly relevant for grass-based roboitc systems. The 'late cow' issue (together with prolonged intervals and low robotic efficiency) would be more pronounced and likely if cows must travel to feed further away from robots (spatial constraint), as the frequency of feeding (fresh grass breaks or PMR feeding) decreases (temporal constraint), and will be seriously aggravated by a combination of both a serve spatial and temporal feeding constraint. To minimize traffic challenges, spatially and over time, robot layouts and feeding platforms may include further options to allow for 2, 3, 4, or even more grass and/or PMR feeding events strategically planned around robots and throughout the day. Below, I share the layout of the MSU's Robotic Dairy Farm (left), which was designed to test a combination of coww traffic and feed options. The layout allows for integrations of robots in a freestall barns (with free cow traffic) and side feedpads, and with optional feeding of pasture with up to a four-way grazing layout (A-B-C-D ) through use of smart-gates. Grazing occurs through a guided cow traffic (milk first system) and cows have the option for free return from pasture through one-way (finger or vertical) gates (right).
In the figure above I show the design for a 4-way sorting system using smart-gates: 1) return to milking, 2) exit platform A, 3) exit Platform B, and 4) exit Platform C. The same design could be expanded further to a 5-way or 6-way sorting on a same location
Already wondering what could be the most convenient robot layout for your farm, or wondering about how a given grazing or feeding plan would impact your robots? Please, do not hesitate to reach me out here. I will be keen to learn about your goals and to share my opinion on your planning, startup or management process