Field Observations on Barn Layout and Design
for Robotic Milking of Dairy Cows

Presented at the Sixth International Dairy Housing Conference, Minneapolis, USA, June 17, 18, 2007

Table of Contents

Abstract. Based on survey results, labor efficiency, convenient one man operation, free cow traffic and cow comfort are among the priorities of robotic milking farms in eastern Canada. When herd management is focused on these priorities the criteria for barn design are different from traditional freestall barns designed for parlor milking. Grouping, training and handling strategies have a major impact on barn layout. Since many questions about ideal management strategies remain unanswered, practical barn layouts that can be readily adapted to different management options are presented. In a separate project two years of experience with a robotic milking dairy located more than 6 hours from technical service support is described. With excess milking capacity, a stock of parts, additional training and satellite video and computer linkages this farm has successfully milked 68 cows without interruption for 28 months using a robotic milking system without the support of a technician.


Since its commercial introduction on an Ontario dairy farm in 1999, an estimated 200 single box robotic milking systems have been installed on an estimated 120 farms throughout North America. Close to 90% of these farms are located in Ontario and Quebec. As part of a larger project, 43 herds on formal milk recording programs in Ontario and Quebec were visited in the summer of 2006. Herd owners were interviewed and details about barn design and management as well as milking performance of the systems were recorded. Although the large numbers of variables in this field data makes it impossible to draw conclusions about the impact of barn layout on performance, the survey data helps to define current practices. In the absence of research information on which to base barn design, field experience gathered through informal contact with these herds and others can be used to identify potential improvements in design to correct current problems and concerns.

Herd Goals for Robotic Milking

Among these 43 herds, the primary reason for choosing robotic milking was the potential for saving labor, mentioned by 27 owners and defined by 18 as specifically "avoiding or reducing the need for non-family help". Other reasons and the frequency they were mentioned included: flexibility and lifestyle (19), a desire to be innovative (9), lower building costs or lower total investment than a parlor (8), increased milking frequency (7), greater cow comfort (5) and minimizing physical work in some cases to relieve operator health concerns (4). These reasons are similar to a field survey (Rodenburg 2001) of the first 15 herds in which "expansion without hiring non-family labor", "improving milk yield through more frequent milking", and "improving lifestyle through greater flexibility in work scheduling", were the most common reasons given for choosing robotic milking. It is noteworthy that "more frequent milking" is cited less often as a benefit of robotic milking today, perhaps because both experience and research (DeKoning 2004) suggest the production and udder health benefits of more frequent milking are much smaller when it is irregular and voluntary than with 3x parlor milking.

Fetching Cows and Free Versus Forced Cow Traffic

In 2001, Canadian owners reported 10 to 15% of the herd did not attend for milking voluntarily and in a subsequent formal study (Rodenburg 2002), 19.2 + 12.5 % of cows were "fetched" by the operator. Owners report the effort required to fetch cows is minimal and normally involves identifying them from milking interval data in the computer, followed by a process of chasing these cows from the freestalls and heading them in the right direction, so that they end up in the holding area. Fetching cows is usually combined with cleaning freestalls. Barn layouts and gating that define a route for stall cleaning and fetching make the actual additional time required for fetching very small. Nevertheless, the need to fetch cows remains one of the main concerns owners have about robotic milking systems. In the current survey, producers reported fetching 14.6 + 10.3 % of cows either once or twice daily. Variation in number of cows fetched was very large. The five best herds averaged 2.5% of cows, while the 5 worst averaged 41.6%. The reasons for fetching cows are listed in table 1. Fetched cows include lame cows, cows with mastitis, and injured animals. When cows first appear on the fetch list, this can assist with early diagnoses of health problems, especially in free traffic herds. As shown, training new heifers and cows is also an ongoing need that must be addressed in barn design.

Table 1. Reported reasons for fetching cows for milking in 43 survey herds


% of fetched cows

% of all cows

Fresh or inexperienced cow (less than 14 days experience)


1.6 + 1.8

Milked manually due to teat placement


0.7 + 1.2

Milked manually for other reasons


0.3 + 1.0

Cow is visibly lame


2.8 + 9.0

Cow has clinical mastitis or teat injury


0.8 + 1.7

No identifiable reason for no-attendance - "lazy cow"


8.5 + 9.9

Total fetched cows


14.6 + 10.3

Herds were defined as having "free cow traffic if cows could move freely between the freestalls and the manger area at one or more points in the barn, or as "forced cow traffic" if cows had to pass through the milking stall or a selection gate in one direction and a one way gate in the other, between these two areas of the barn. Cows fetched in the 35 herds with some form of free cow traffic was 16.2 + 10.8% and this was much higher than the 8.52 + 5.9 % in the 8 forced cow traffic herds. The merits of free and forced traffic have been reviewed (Rodenburg 2004). Both free cow traffic and forced traffic with pre-selection appear to give reasonably satisfactory results. Where high feed intake and productivity per cow is desired, free cow traffic should be the preferred cow routing system despite the fact this results in the need to fetch a larger number of cows. In circumstances where minimizing the labor of fetching cows is highly valued, the continual use of forced cow traffic with pre-selection gates, preferably at remote crossing points is recommended. Despite greater fetching labor, the majority of robot owners in this survey chose free cow traffic. The main reasons given for this were improved cow welfare and the expectation of higher production. Practical designs for free cow traffic barns should encourage the maximum number of cows to attend voluntarily, while making fetching and training convenient when it is required.

Consideration should be given to the fact that herd dynamics changes over time. For example, field data (Rodenburg 2002) suggests with experience, the number of fetched cows is likely to decrease and the frequency of voluntary milking increases. As animals selected for their willingness to attend voluntarily begin to populate the herds over time, the labor of fetching decreases. This may make larger groups of cows accessing multiple robots more practical. Forced traffic barns will increasingly be populated by cows eligible to pass through selection gates at will. Since many of these barns are currently built with 4 rows of freestalls accessing one manger area, manger space is restricted to 1 foot per cow. This works today because only half the cows are eligible to pass through selection gates, but in the future, limited manger space will result in crowding and competition at the manger when fresh feed is delivered to the barn.

The Need for Service Support 

Historically, robotic milking technology has only been sold commercially in areas where it can be supported by professional technicians able to provide emergency service on a maximum two hours notice. This dependency on professional technician service likely evolved because regular milking is a requirement for dairy cows; because early commercial systems did have high failure rates; and because farmers' perceptions of the complexity of robotics makes them wary of providing even a basic level of service themselves. In Western Europe where the geographic dairy farm density is high, this service model works quite effectively but this requirement has slowed the introduction of robotic milking in many parts of North America. Northern Ontario is a large geographic region with small pockets of dairy production, and it is an area where 2 hour technician service for robotic milking systems cannot possibly be achieved. To determine if robotic milking could be supported in a different fashion in more remote areas, the Ontario Ministry of Agriculture, Food and Rural Affairs initiated a demonstration project, in which 2 Lely robotic milking stalls were installed on Jonella Farms located near the town of Massey, in northern Ontario, approximately 6 hours by road from the nearest service technician. Prior to installation, the owner was given two weeks of service training. A buffer of excess capacity was provided by installing two milking boxes to milk 68 cows, providing nearly double the normal capacity. A stock of parts normally carried on service trucks with an estimated value of $40,000 was also placed on the farm. To provide professional technical support, audio communication by cell phone was supplemented with a video satellite link using cameras in each robot room, as well as remote software management capabilities by satellite. A farm with two identical robots serviced under a typical contract was chosen as a control herd. Since this farm did not have excess capacity these two robots milked an average of 112 cows instead of 68, so a somewhat greater service requirement would have been expected. Since the start of robotic milking at the northern Ontario site in Dec 2004, the longest interruption in milking in more than two years of operation has been 1 hour, and all technical problems have been resolved successfully with remote technical support. The excess capacity has been essential. On one occasion the farm operated on a single milking stall for 11 hours due to a milk pump repair and for 16 hours on another occasion involving a faulty computer card. This requirement for additional capacity and for stocked parts implies robotic milking in remote areas probably remains uneconomical due to these extra costs. The owner reports on four occasions in the last two years, he felt he would have benefited substantially from having emergency service close at hand. At 30 to100 Kb/sec, the satellite video and computer link is often frustratingly slow for live data transfer and video. It has been used successfully to install new software, a function for which speed is not as critical, but as a video link it has been unreliable at best. In general the excellent performance of the equipment itself and the diligence of the owner operator has made this a satisfactory experience and the farm continues to milk with robotics today. Service and maintenance records for the first 12 months of operation for this farm and the control herd are summarized in table 2.

Table 2. Service and maintenance summary for a remote location robotic milking herd vs a control herd with traditional technician service. Dollar values are Canadian Dollars.

Remote service herd

Control herd

Number of cows/number of milking robots



Contracted maintenance fee (package)

$7775/yr (Standard)

$14700/yr (Standard plus)

Parts supplied during maintenance visits

$1655 (excluded in contract)

$3944 (included in contract)

Parts shipped to replace on-farm stock

$687 (excluded in contract)


Scheduled technician visits to the farm

4 visits, total 24 hrs. plus 44 hrs travel time

5 visits, total 17 hrs plus 6.25 hrs travel time

Unscheduled technician visits to the farm


16 visits total 32 hrs plus 20 hrs travel time

Phone calls into service line



Total cost to producer

$10117 plus his own maintenance labor


Total technician time invested in service:

Including travel

Excluding travel

68 hrs

24 hrs

75.25 hrs

49 hrs

* information concerning length of these calls was not available but was undoubtedly much longer for the remote service herd.

The number of unscheduled visits to the control herd is well above the company average of 6 per year, however it appears that when the remote location demanded greater diligence from the operator, the number of true emergencies may be as low as 1 or 2 per year. Perhaps the established use of full service contracts, which are offered to address the perceived concern about high service requirements, is actually self fulfilling in that it promotes dependency on service and encourages a disregard for simple daily care and maintenance on the part of the owner. Our experience in Northern Ontario suggests it is time to reassess the "full service culture" that has evolved with robotic milking and to develop other service models which result in fewer service calls and lower service costs.

Unanswered Questions Concerning the Design of Robotic Milking Barns

Common barn designs in current use have been described (Rodenburg 2004, Choini�re 2007). These designs have been based on a combination of very limited research, limited practical experience and theoretical and perceived notions of cow behaviour and the needs of the operator for management. Some aspects of robotic milking are clearly different than parlor milking and need to be addressed differently in robotic milking barns. Since cows never leave the barn, cleaning alleys with mechanical scrapers or with flushing or with slatted floors is much less intrusive than tractor scraping, and nearly all barns are built to accommodate one of these options. Because cows are never routed to a parlor, cow movement from group to group through the lactation cycle can also be complicated if chronological groups are no housed adjacent to each other. Many other aspects of ideal management for robotic milking are as yet unresolved. Some of the remaining questions include:

  1. In a barn with two or more milking stalls, is it best to manage the barn as individual groups of 60 cows, each accessing one robot, or as one group of 120 or 180 cows with access to two or three robots? Experience has shown finding and fetching a smaller number of cows from several 60 cow groups takes less time then finding a larger group of fetch cows among 120 or 180 cows, but larger groups result in barns with fewer gates reducing labor associated with other tasks. Larger groups accessing multiple robots would also reduce waiting times, especially for aggressive cows and during times of washing or maintenance service. If a heavily used single box is down for repair for more than 2 to 3 hours, owners report production losses can be substantial. These losses are less severe when groups can access two milking stalls. At start up, 60 cow groups may be desirable but an ideal barn layout should permit the use of larger groups a few years later. Of the 43 survey herds, 12 provide access to more than one robot by a larger group of cows. Owners of these herds are convinced they benefit through higher milk production, but due to the large number of variables our field data cannot corroborate this. Data on milking frequency and the extent to which cows use multiple robots when access to more than one is offered has been collected and will be analyzed in 2007 to determine if specific configurations for grouped milking stalls influences the likelihood that cows will select the same milking stall at each milking or choose any available milking stall..

  2. If cows are grouped in groups of 60 and there are two milking stalls, is it advantageous to split the herd by stage of lactation and move cows from a high group to a low group? Owners report diets high in grain decrease milking frequency; grain in the milking box is an important factor in attracting cows for milking and late lactation, lower producing cows are most likely to require fetching. If a herd can be grouped by stage of lactation, the grain level in the TMR can be reduced in mid lactation to permit greater grain feeding in the milking stall, and hence better voluntary attendance by late lactation cows. Of the 43 survey herds, one currently splits the herd this way and one has tried and abandoned this practice. The two robotic milking stalls in this herd where a right and left hand model and approximately 10% of cows had a very difficult time adjusting to the different entry point. Grouping by stage of lactation has possible negative impacts if the group change increases stress due to adjustment to a new group of herd mates. Barns that have the option to be split in this way, with convenient movement between groups would be an asset if future experience suggests grouping by stage of lactation is desirable.

  3. Is it beneficial to give cows and or heifers access to a milking stall prior to calving? More than 10% of fetched cows are heifers and cows being trained to the system, and since these animals usually require greater effort than cows fetched for other reasons, training new cows involves a substantial percentage of the labor associated with robotic milking. Eleven of the 43 herds surveyed currently introduce heifers to the robotic milking stall prior to calving and provide training to use the stall by offering feed in the stall. Some of these herds also introduce arm movements prior to calving to maximize the pre-calving experience and others lead feed dry cows in the stall as well. In 9 of the 11 herds, training cows prior to calving involves putting them in the milking group, a decision which alters the ration and makes separating the cow at calving more difficult. One negative outcome of using a milking stall for training is it reduces its capacity for milking cows. As labor costs increase and the relative cost of investment in robotic milking stalls decrease, the dynamics of pre-calving training and lead feeding in the robot also improves. Ideal barns will have the capability to address this need if the perceived benefits increase the adoption of this practice.

  4. Is it beneficial to maintain a separate group of fresh and or lame cows on a straw pack, and if yes do these cows need free access to the milking stall? In recent years, many parlor milked herds have added a small separate group of milking cows consisting of fresh and lame cows housed in a minimal stress area of large freestalls or a bedding pack. By reducing competition with more aggressive cows and reducing the waiting time for parlor milking, these dairies report that fresh cows get a better start and lame cows recover more quickly. In a robotic herd, lame cows seldom attend voluntarily and finding and fetching them requires much less labor if they are in a separate group.

  5. Do cows exiting a robot need to be oriented towards the feed manger and if yes do cows need to be trained to use both a left and right exit robot if they will use both during their lifetime? Most traditional layouts of robotic milking barns assume cows must enter from a resting area and exit facing the feed manger, and in barns with multiple robots this usually leads to including both left and right entry stalls. Field reports suggest changing a cow from one entry side to the other in mid lactation is a problem for 10 to 15% of cows and a smaller number require retraining if changed between lactations. No studies have been conducted to determine if placement of the milking box with its entry on the feed alley side actually reduces milking frequency or harms cow flow. In fact with forced cow traffic, DeLaval currently promotes a "feed first" traffic pattern as desirable. The possibility that barns should be designed with only left or right hand entry stalls to reduce training times and increase milking frequency should be investigated.

  6. Is separating cows at the time of milking a practical method of sorting and handling in robotic milking systems? In Eastern Canada, most robotic milking barns have the capability for post milking separation, but since milking time cannot be predicted, work scheduling usually dictates cows are either handled in the freestalls or in head locks. In herds where cows are sorted in the milking stall, it is usually after they are fetched into the holding area, refused for milking and then sorted in the sort pen, rather than being sorted after a voluntary milking. Separation pens that can combine this short term handling function with alternative housing for cows from question 4 would be ideal. In free traffic robotic milking barns, headlocks are also a valuable tool for group handling such as pregnancy examinations etc. In the future, it may be practical to design headlocks that can selectively release identified animals. Such a system could lock up all animals at feeding, then selectively releases long milking interval cows as an alternative to fetching. After the cows to be milked are chased to the holding area, all cows except those needed for handling could be released. Handling work could be scheduled to follow this step, so no labor is expended looking for cows. If headlocks are a valuable tool in robotic milking barns, the trend to forced cow traffic designs with four freestall rows along one manger could prove detrimental.

Flexible Barn Designs for Robotic Milking

Undoubtedly answers to many of the above questions will become clearer as additional research is conducted and as producers gain additional experience with robotic milking. Since there are so many unanswered questions, and since the barns we build today will be in use for several decades, flexibility in design, is an important attribute for robotic milking barns. Fig. 1 and Fig. 2 are presented as examples of expandable facilities for 120 cows incorporating the flexibility to work with a number of different grouping strategies. They offer the ability to choose among the priorities listed above simply by adding or moving gates with no changes to the barn itself. Other challenges unique to robotic milking facilities are also addressed by these designs and are discussed. These plans were developed using the experiences of advisors working with robotic milking herds in Europe as well as North America.

Fig. 1 Flexible Robotic Milking Barn for 2 milking stalls, and milking and dry cows. Milking Stalls exit to the feed manger.

Robotic Milking Barn for 2 milking stalls, and milking and dry cows.   Milking stalls exit to the feed manger.

Both of the plans include 120 freestalls in 6 rows with drive through feed alleys on the outsides of the barn. This layout allows the cows in the main freestall area to be handled as one group with no gates or as two groups with gates at both ends of the center double row of stalls. The difference between the two plans presented is that Fig. 1 has right and left entry milking stalls facing the feed manger on each side of the barn and Fig. 2 has robots facing the same way. Industry nomenclature on whether these two robots are left or right hand models is currently inconsistent. Lely calls the robots in Fig.2 right hand stalls because the arm approaches the cow from her right side. DeLaval calls then left entry stalls because the cow is to the left of the arm as she would be to the left of the operator in a parlor situation. Since it is known cows have difficulty adjusting to different entry sides, and since the need for exit toward feed is being questioned today, Fig. 2 may be the better choice. Ontario field experience suggests a large open area in front of the milking stalls enhances cow movement in the barn. This open area also makes it easier for cows to negotiate an exit route to either resting or feeding areas regardless of the orientation of the milking stall.

With 4 foot wide freestalls, a 12 foot crossover at the end of the barn and a 20 foot open area in front of the milking stalls, these plans provide 2 feet of manger space per cow for 56 cows per side along the drive through feed alleys at A. Since ideal handling methods for these barns are poorly defined and since headlocks may well play a role in future management schemes, it is tempting to widen the crossover an additional 4 feet to provide space for one headlock per freestall. Feeding from two drive through alleys, on the outside of the barn is more work than a center drive through but makes it possible to handle the herd as one group and makes cow movement between groups and work among the cows more convenient. The outside drive alleys also keep rain and sun out of the cow area. Narrow cross alleys B at the ends of the barn permit workers and small equipment to travel around the perimeter of the cow area to push up feed. If alley scrapers are used, manure drops at the end farthest from the milking stalls and a place to "park" scrapers out of the traffic area can be incorporated under this alley. If slatted floors are used, the 24 foot post spacing is compatible with under the barn manure storage.

Fig. 2 Flexible Robotic Milking Barn for 2 milking stalls, and milking and dry cows. Both milking stalls exit in the same direction

Figure 2: Flexible Robotic Milking Barn for 2 milking stalls, and milking and dry cows.  Both milking stalls exit in the same direction.

Gates at C and D can be used to divide the herd into two 60 cow groups. To fetch cows into holding area E, close gates C and F and clean the freestalls in a counter clockwise direction, keeping fetched cows ahead of you. Close gate D when you come to it and complete cleaning stalls in this half of the barn. Repeat this in the other half of the barn, by closing gate G and opening D as you pass it. Once the fetched cows are in the holding area, all gates can be opened. Since the holding areas E and H are only used to house fetched cows, access to the milking boxes for other cows via the "split entry" is unrestricted. High ranking cows from the main milking groups are kept out of the holding areas reducing stress for the usually more timid fetched cows. Note there is a gate in the holding area behind each milking stall which can be used to direct and "squeeze" an inexperienced cow into the milking stal.

A bedding pack for fresh and lame cows at I has access to the milking stall beside it through holding area C in Fig. 1 and F in Fig. 2. With a post milking sort gate, these cows are returned to the pack. Far off dry cows housed in the freestalls in the center area J have manger access beside the cows in the bedding pack. A movable gate in the interior alley at K separates far off dry cows from close up cows in area L, which eat at the manger behind the other milking stall. The close up cows have free access to the milking stall beside them for lead feeding and training purposes, and a post milking selection gate will return them to their pen area. At calving, close up cows can be easily moved into one of three bedded pens in area M.

Clean access to the milking stalls is via a bridge with a 36 inch gap at N .To segregate and restrain a cow for individual handling, she can be fetched into the holding area, and post selected through the milking stall. If the bridge at L is gated, the segregated cow can be restrained in a headlock at O, treated and released back to the main barn. Handling and especially hoof trimming could also be incorporated at P, a spot that is readily accessible to all cow groups. Cows can be easily moved from group to group through a lactation cycle, and all feeding is along the two mangers on the sides of the barn. The bedding pack and calving pens are accessible from the end of the barn for clean out. The office is located to give a good view of both the area in front of the robots and the calving area. With a sliding window at Q and a turntable for the computer screen and keyboard, clean and dirty access to the computer can be provided. In Fig. 1, adding a lane across the width of the barn behind the robots would make it possible to train close up cows on both a left and right entry, or allow the cows housed in the pack access to both robots if that was preferred. In this case only one of the groups would have ongoing access while the other could be moved to the milking stalls by resetting the gates and fetching the group.

The principles demonstrated in these plans can also be applied to larger herds. To double the herd size, three options are suggested. It is possible to mirror image the barn on the end used for dry cows and create a barn with 4 robot rooms located at the corners of a central handling area. Moving far off dry cows to another facility would keep the central handling facility more compact. This option leaves all the grouping possibilities intact, but it does require clean out of the bedded areas from the side of the barn. Back to back milking stalls with no separation and handling facilities, combined with another barn with at least one milking stall set up for training and handling is a second option. If the choice is made to work with a minimum group size of two milking stalls and 120 cow groups, Fig 3 provides a configuration combining two same sided milking stalls. In Fig. 4 this configuration is used to double the length of each end of the barn, and add a single robot room with two back to back milking stalls.

Fig. 3 An ideal layout for two milking stalls in a single group of cows. Fetched cows collected in holding area HA are milked in milking stall 1. Separation cows milked in stall 1 are sorted into Separation area SA, and can access the holding area and milking stall via a lane behind the robot room. Cows separated in milking stall 2 sort into the holding area, are subsequently refused in milking stall 1 and directed into SA.

Figure 3: An ideal layout for two milking stalls in a single group of cows.

Fig. 4 An expansion of the barn in Fig. 2 to include 2 groups of cows each accessing two milking stalls

Figure 4: An expansion of the barn in figure 2 to include 2 groups of cows each accessing two milking stalls.

A forced traffic version of the layout in Fig.1 would also be possible by eliminating the outside rows of freestalls on each side and lengthening the barn to accommodate 120 stalls.


Undoubtedly robotic milking technology will continue to evolve, placing further demands on barn design that have yet to be considered. Producers in eastern Canada that have adopted robotic milking indicate labor saving, and a flexible lifestyle are the primary reasons for choosing robotics, but even though free cow traffic increases fetching labor the majority have chosen this management system. Hence future concepts of design should emphasize labor saving and convenient handling in a free traffic setting as well as flexibility to accommodate new grouping and management strategies. Barns that do this successfully will have more open space near the milking stalls, separate or temporary holding areas, adequate manger space for headlocks, convenient facilities for dry and special needs cows, and the ability to make either small or large groups accessing one or two milking stalls. Given that many questions are yet to be answered, giving priority to flexible designs that can accommodate a number of variations in management style is an important consideration in barn design for robotic milking.


Choini�re, Y, and C. Lemay. 2006. Optimization of cow management and sorting capabilities with robotic milking: the Eastern Canada experience. Proceedings of the Sixth International Dairy Housing Conference, in press.

DeKoning, K., and J. Rodenburg. 2004. Automatic Milking: State of the Art in Europe and North America. "Automatic Milking - A better understanding" Proc. of the International Symposium, Lelystad March 24-26 2004, 27-37, Wageningen Academic Publishers

Rodenburg, J. and D.F. Kelton, 2001. Automatic Milking Systems in North America: Issues and Challenges Unique to Ontario. Proceedings of the 40th Annual Meeting of the National Mastitis Council, pp 162-169.

Rodenburg, J. and B. Wheeler, 2002. Strategies for Incorporating Robotic Milking into North American Herd Management. Proceedings of the First North American Conference on Robotic Milking, March 20-22, 2002, Toronto Canada, Wageningen Press, pp III 18-III 32.

Rodenburg, J. 2004. Housing considerations for robotic milking. ASAE/CSAE meeting, Ottawa, paper # 2004-044189.

For more information:
Toll Free: 1-877-424-1300