Introduction
Automation can be defined as technology that allows for a process to be performed automatically, with minimal human input. The adoption and use of automation in the dairy industry has grown dramatically in the past decade. Much of the automation adopted has been associated with milking and other aspects of barn management. There is also growing interest and adoption of automation in feeding of dairy cattle. Specifically, the use of automation in feeding of lactating dairy cows is growing, being necessitated to improve the accuracy and precision of diets fed to dairy cows in effort to improve health, production, and efficiency, as well as to minimize the requirements for human labor input. Human labor for dairy operations may not only be increasingly hard to come by and expensive, but may also be more prone to error and inconsistency in comparison to automated technologies. Further, the rapid adoption of technologies that allow for individualized feeding, including automated milking and calf feeding technologies, has also increased our ability to feed dairy cattle according to their individual requirements, thereby facilitating precision nutrition. This paper will review the various needs for automation in feeding dairy cattle to achieve precision, outlining both opportunities and challenges.
Automation in Dietary Preparation
Improved consistency, or precision, of diets provided to dairy cows may have significant impacts on herd-level performance. Measures to reduce ration variability, such as implementation of automation in diet preparation, may lead to improvements in herd performance and profitability. The need for automation is highlighted by the fact that, despite best efforts, the delivered ration on many dairy farms does not accurately match that which was formulated for the cows. In previous research we observed that as the variability between the ration offered to the cows and the original formulated ration becomes greater, so does the chance that cows will not perform to expectation (Sova et al., 2014). While most of us have always suspected that cows do not always receive the ration exactly as it is formulated for them, this research is some of the first to support this idea and identify the potential consequences of such deviations. For our study we sampled the mixed and delivered total mixed ration (TMR) for 22 free-stall, parlor-milked herds for 7 consecutive days in the winter and summer months. The nutrient analysis of these feed samples was then compared to that formulated on paper for those farms. Across farms, the average TMR fed did not accurately represent that formulated by the nutritionist. The average TMR delivered exceeded TMR formulation for net energy of lactation (NEL), non-fibre carbohydrate (NFC), acid detergent fibre (ADF), calcium, phosphorus, magnesium and potassium, and underfed crude protein (CP), neutral detergent fibre (NDF) and sodium. Theoretically, some deviation might not be hugely problematic because a safety margin is generally included in formulation to account for uncertainty in ingredient composition. Across farms, however, there was a huge range in this variation, with some farms consistently experiencing a 5–10% discrepancy (both positive and negative) between the fed and formulated ration for nearly all nutrients. This lack of dietary accuracy can be costly, either in excessive provision of expensive ingredients, or not allowing cows to meet their production expectations due to nutrient shortfall.
In that study we also investigated the day-to-day consistency in physical and chemical composition of TMR and associations of this variability with measures of productivity. Greatest day-to-day variability was observed for refusal rate, particle size distribution, and trace mineral content. Delivery of a more consistent ration was associated with improved production. For example, greater herd-average dry matter intake (DMI; Figure 1), milk yield (Figure 2), and efficiency of milk production were all associated with less daily variability in energy content of the ration (Sova et al., 2014).
Figure 1. Association between fed ration coefficient of variation (CV) in NEL and average DMI. Coefficient of variation was calculated as the standard deviation of NEL over 7 d divided by the average NEL over 7 d. Figure adapted from Sova et al. (2014).
Figure 2. Association between fed ration coefficient of variation (CV) in NEL and test-day milk. Coefficient of variation was calculated as the standard deviation of NEL over 7 d divided by the average NEL over 7 d. Figure adapted from Sova et al. (2014).
Lower day-to-day variation in the percentage of long forage particles in the offered TMR was also associated with greater milk yield and efficiency of milk production (Sova et al, 2014). On average, day-to-day variability was greater for physical characteristics (i.e., particle size distribution) of the ration compared with the ration’s nutritional composition. This suggests that this day-to-day variation may have been caused by variability in feed component nutrient and dry matter (DM) composition, but probably even more so by mixing errors associated with operators (timing, sequencing) or equipment. Regardless, these findings suggest that increased surveillance of the TMR composition, in addition to individual feed ingredients (e.g., regular, frequent forage DM determination, regular nutrient testing of feeds), may be helpful as a regular component of feeding management to ensure delivery of TMR with the intended nutrient composition to maintain production and feed intake. Further, these findings reinforce the need for standard feeding protocols and training to achieve those protocols.
It is noteworthy that all of these practices to ensure dietary accuracy and precision may be enhanced through use of automated technologies. For example, real time sensors (e.g. NIRS) can be used on feed preparation equipment to determine dry matter and nutrient content of feed ingredients being incorporated into diets. TMR management programs may be used for more accurate weighing of ingredients, easier adjustment and calculation of pen delivery amounts, and monitoring of mix consistency (both timing and composition). Finally, automated feeding systems are available, which may be setup to fully automate the collection of dietary components and mixing of those together into a TMR.
Automation in Diet Delivery and Feed Access
Automation may also be important for delivery of diets, as well as in maintaining consistent and continual access to feed, which are key components in maintaining precision of nutrient consumption. The diurnal feeding patterns of dairy cows fed a TMR are primarily influenced by the time of feed delivery, feed push-up, and milking (DeVries et al., 2003). Of these, the delivery of fresh TMR has the single largest impact on stimulating feeding activity at the bunk (DeVries and von Keyserlingk, 2005; King et al., 2016a). As a result, greater frequency of feed delivery can greatly influence feeding behavior patterns, promoting more consistency in feed activity across the day (DeVries et al., 2005). In some studies, greater frequency of TMR delivery has also been associated with greater DMI (Sova et al., 2013; Hart et al., 2014). Further, delivering a TMR 2x/d or more often has also been demonstrated to reduce the amount of feed sorting compared with feeding 1x/d (DeVries et al., 2005; Endres and Espejo, 2010; Sova et al., 2013), which would further contribute to more consistent nutrient intakes over the course of the day. Such desirable feeding patterns are conducive to more consistent rumen pH, which likely contributes to improved milk fat (Rottman et al., 2014). In support of that, Woolpert et al. (2017) reported that dairy herds with high de novo fatty acid concentration in bulk tank milk, compared with those with low de novo fatty acid concentration, tended to be 5x more likely to be fed 2x versus 1x per day, confirming the positive impacts of feeding > 1x/d on maintaining a consistent rumen environment.
Implementation of greater TMR delivery frequency on dairies is often constrained by time and cost associated with TMR preparation and its delivery. Thus, implementation of feeding automation, not only for diet preparation, but also for frequent delivery to cows across the day may have significant benefits in terms of achieving greater precision. There is, however, a paucity in research on the effectiveness of the implementation of these automated TMR mixing and delivery systems.
TMR push-up is also critical to ensure that feed is accessible when cows want to eat. Feed push up needs to occur frequently enough such that any time a cow decides to go to the feed bunk, there is feed available to her. Feed push-up also helps minimize variation in feed consumed because it mixes up the feed that is no longer in reach with that which is currently available in the bunk. Thus, frequent pushing up of TMR in the bunk is necessary, particularly in the first few hours after feed delivery, when the bulk of the feeding activity has occurred. We have demonstrated that greater lying duration is associated with greater frequency of feed push-ups (Deming et al., 2013; King et al., 2016b), suggesting that frequent push-up minimizes the time cows need to spend waiting for feed access and cows can devote more time to lying down. Feed push-up will also ensure that DMI is not limited and thus production is optimized. Evidence for this was shown in a cross-sectional study of 47 herds, all with similar genetics and feeding the exact same TMR (Bach et al., 2008). In that study it was reported that those herds where feed was not pushed up (5 out of 47 herds) produced 3.9 kg/d/cow less milk (-13% difference) than herds where feed was pushed up. Interestingly, in a more recent observational study of robot herds, Siewert et al. (2018) reported that farms with automatic feed push-up produced 352 kg more milk/robotic unit and 4.9 kg more milk/cow per day than farms that manually pushed up feed. This effect may not be directly attributable to the use of an automated feed pusher, but rather that those farms using such automated equipment had more consistent feed push-up, and thus continuous feed access, than those pushing up feed manually. In support of this, in a recent observational study of 197 robot herds in Canada, we demonstrated that each additional 5 feed push-ups per day was an associated 0.35 kg/d greater milk yield at a herd level (Matson et al., 2021). Thus, in situations where manual feed push-up is done consistently and frequently, the same results should be achievable. Unfortunately, in reality, manual feed push-up, performed by farm staff, is more prone to inconsistency, in time and frequency; thus, this again provides supports for the use of automation.
Automation in Individualized Feeding
In addition to herd-level feeding precision achieved through automation, there is also opportunity to use automation to achieve that precision at an individual animal level. The rapid adoption of technologies that allow for individualized feeding, including automated milking and calf feeding systems, has also increased our potential ability to feed cattle according to their individual requirements.
Given the ability to supplement the feed consumption of dairy cows within automated milking systems (AMS), there is potential for applying some type of precision feeding approach in AMS (Bach and Cabrera, 2017). While there is potential, there are many challenges with such an approach. In AMS, the herd is fed a common diet (partial mixed ration – PMR) at the feed bunk. As this PMR diet is static, the ‘precision’ aspect, to meet individual cow nutrient needs, would need to be accomplished with the feed provided at the AMS. In theory, if the individual nutrient requirements were known (based on expected milk production, and other known factors including age, body weight, stage of lactation, pregnancy status) then the amount of feed provided in the AMS could be adjusted to their individual need. The challenge with that is then also be able to accurate predict the nutrient consumption from the PMR, as that is not measurable on an individual basis in commercial settings. Therein lies some difficulty, it has been demonstrated that the level of PMR consumption is affected by the level of concentrate provided at the AMS, and it is not necessarily an even substitution ratio (Hare et al., 2018; Menajovsky et al., 2018; Paddick et al., 2019; Schwanke et al., 2019). In fact, across studies, the substitution rate (amount of decrease in PMR intake for every 1 kg increase in AMS pellet intake) has ranged from 0.63 to 1.58 kg (Hare et al., 2018; Schwanke et al., 2019). So, it may be difficult to predict total DMI, and thus total nutrient intake, when varying the amount of feed provided at the AMS, making precision feeding more difficult. Further, in studies where we have increased the quantity of AMS pellet offered in the AMS, the day-to-day variability in the consumption of the AMS pellet also increased (Hare et al., 2018; Menajovsky et al., 2018; Paddick et al., 2019; Schwanke et al., 2019). This variation then makes the concept of precision more difficult to attain. A further challenge with feeding in AMS is that just because cows are provided feed at the AMS, does not guarantee they will consume it (Bach and Cabrera, 2017). Any unconsumed feed left in the AMS results in another cow potentially consuming more than what they are programmed for; this reduces the ability to precision feed these animals.
While most AMS are only equipped with a single bin for delivering concentrate to cows (Bach and Cabrera, 2017), there is opportunity within many systems to provide multiple feeds. It possible that greater precision in feeding could be achieved in such scenarios, as the amount and balance of different types of supplement feeds could be used to match individual cow nutrient requirements. To date, however, there is limited research on this type of approach. In a recent study, we demonstrated that we could improve energy balance and minimize body condition loss in early lactation by supplementing cows milked in AMS with a molasses-based liquid feed supplement in addition to their regular AMS concentrate (Moore et al., 2020).
There may be opportunities to apply such precision feeding principles in other types of milking systems. One such example is that described by Bach (2014), a ‘dynamic concentrate parlor feeder’ which involves the preparation and delivery (in real time) of many different feeds (in both quantity and composition) within a rotary milking parlor. The system calculates the individual nutritional requirements of each cow entering the parlor based on her assigned feed intake (average of the pen where she is), composition of the TMR fed, DIM, parity, BW, BW change, pregnancy status, milk yield, and milk component yields. Based on those needs, the system creates a least-cost formula using up to 6 feeds that are mixed and delivered to the cow in less than 14 seconds. Bach (2014) suggested that such a system would allow for the feeding of a more cost-effective TMR with a low nutrient density, without compromising, and even potentially improving, income over feed cost by delivering nutrients to only those cows in need of them.
One area where there has been more success in application of individualized feeding strategies is with the use of automated calf feeders. Automated calf feeders provide the ability to feed calves individualized diets that may be calf-specific based on age, weight, or any other parameter deemed appropriate. This may include altering the speed at which milk allowance is increased in early life, as well as decreased at the time of weaning. To date, however, much application of these feeding strategies, while applied at the calf level, is still done similarly across all animals within a farm. There is research to suggest that much gain can be made by tailoring feeding programs for individual calves based on their individual needs. For example, de Passille and Rushen (2016) demonstrated that individual calves differ greatly in when they begin to consume solid feed and how quickly they increase the intake in response to a decrease in milk allowance. Those researchers demonstrated that automated milk feeders could be used to wean calves at variable ages, depending on their ability and willingness to eat solid feed.
Conclusions
Automated technologies have been developed and increasingly adopted within the dairy industry to not only reduce human labour requirements, but also to increase the accuracy and precision of application of various management tasks. Various forms of feeding technologies are currently available to increase our precision in feeding strategies of dairy cattle. At a herd-level, this includes automated feed preparation and delivery. While, at the animal level, this includes individualized feeding opportunities, to date primarily through automated milking in lactating cows and automated milk feeders in calves. While there are still many challenges associated with the successful implementation of such precision feeding strategies, on-going research would suggest that these opportunities will continue to grow, allowing for greater nutrient capture, greater efficiency, less nutrient waste, and greater health and production.
Acknowledgements
Much of the research presented in this paper was funded by the Natural Sciences and Engineering Research Council of Canada, Dairy Farmers of Canada, Agriculture and Agri-Food Canada, the Canadian Dairy Commission, Dairy Farmers of Ontario, the Ontario Ministry of Agriculture, Food, and Rural Affairs, Eastgen, the Canadian Foundation for Innovation, the Ontario Research Fund, and the University of Guelph.
Presented at the 2021 Animal Nutrition Conference of Canada. For information on the next edition, click here. 
