Large animal - September 2026

The automatic milking system (AMS) has existed as a concept since the 1980s and has been commercially available since 1995. Automatic milking systems are now the fastest growing segment of the global milk harvesting market. Not limited to milking, robotic solutions are now seen in feeding, manure management, footbathing and even scratching brushes. Due to sensors and computerised detection systems, the amount of data and analytical opportunities coming from AMS technology has brought innovative insights beyond the capabilities of more conventional setups. Milking an individual teat rather than applying vacuum to all four until take-off is a prime example of this optimisation. This article will concentrate on milk harvesting but aims to touch on the advantages and challenges of feeding and manure robotic solutions.

Barn design and management systems for an AMS

Grazing in a voluntary AMS can be very successful and is increasing in popularity: the pasture can be separated into two or three areas – A, B and C – and cows exiting the robot are sorted through a gate that changes every eight hours (three times a day). C-area may be a shed for buffer feeding, or a smaller paddock and used overnight. Strips within each area provide the grass energy for the allocated number of cows for eight hours and are moved daily. Good quality cow tracks are essential in these systems, with water points provided along the track. Optimal mobility management is critical in these herds.

The large differences between AMS manufacturers’ recommendations for new design lie in the choice between guided or free-flow cow traffic. Guided systems require cows to be collected manually to a collecting area through gates and to be brought to the robot; other options include a one-way system through the barn or a free cow voluntary traffic system. In the latter system, whether indoor or pasture-based, the incentive to visit the robot is by concentrate feed offered, or fresh pasture after visiting. There have been various publications comparing the two systems (Cutress D., 2020) (Cook A., 2023). Guided systems tend to require more labour but may allow more cows per robot whereas free cow traffic centres around cow choices and requires cow mobility to be optimised for peak performance. It is essential to adjust milk access to limit waiting times (Solano L., 2023).

Whether new build or retrofitting, various considerations are recommended in indoor setups:

at least 5m of distance between robot and stalls, providing 10cm/cow linear access of water in this area and throughout the barn;

lighting above the robot area will encourage voluntary visits around the clock and cooling fans in collection areas are recommended; and,

other types of robots can be utilised to clean flooring, offering more flexibility than traditional systems - these can be programmed to run more frequently in areas resulting in less risk of bow-waves, allowing cows to step around the machines rather than stepping over an auto-scraper.

Robots and feeding

The consistent approach from feeding robots can help with dry matter intakes by targeting fresh feedout times and push-back times (DeVries T., 2019). The robots can feed through dropping from a rail above, automatically driving, mixing and depositing along the feed aisle, or scanning for feed fence height and feeding fresh feed continually to demand. If properly set up and maintained, this can reduce the risk of empty feed fences and improve dry matter intakes over 24 hours as well as reduce the use of fossil fuels. More consistent and frequent feedouts can mitigate the risk of sub-acute ruminal acidosis in dairy cows (Humer E., 2018).

The human factor: management and routines

Robots may free up time spent in menial tasks, which gives more time through the day to spend on value-added gains such as observation of the herd, machine maintenance, proactive hoof health, and analysis of data to optimise performance and health. Washing down components, unblocking bleed-holes, and checking cows are essential daily tasks. Replacing liners, brushes, and milk lines will ensure parts last longer and udder health is maintained. Servicing and maintenance become very important in an AMS so ensuring there is a robust protocol for booking can prevent breakdowns, which can be detrimental to performance. Regular reviews with the manufacturers’ on-farm advisors alongside veterinary and nutritional advisors are recommended to reduce subsequent health problems in cows and optimise the system. Preparing new-start farmers with adequate training before installation is an excellent way to prevent transition stresses in moving from a conventional system to an AMS.

The milking process

The aim of milk harvesting on any system is to milk every teat quickly, completely, and gently at every milking moment. Whether or not this is achieved is influenced by many factors involving barn, cow, and machine management. Most robots differ in specifics but the timeline of cows entering to leaving the unit are similar in principle. Cows enter through a gated system, usually designed to reduce risks of social bullying from others during entry. The cow is identified using an electronic tag and/or camera photocell, and is either closed in for milking or released if her defined interval between milkings has not yet been reached. She usually has feed delivered relative to her daily production and shortly afterwards the teats will be prepared. Some units use a water flushing and pulse system to stimulate and clean the teat and other use a disinfectant-washed, spinning brush. Both rely on touch-stimulation to coincide with let-down by the time attachment occurs. Stimulation settings are very important for maintaining a consistent peak milk flow: insufficient total brush touch time often results in bimodal milking (Figure 7). Some systems can use past milking flow profiles to predict the optimum amount of brush time for each teat and adjust automatically. Bear in mind that this can affect total time that the robot is free in 24 hours. Teat detection may be performed using weight distribution (older technology), camera, or laser. Accuracy has increased with innovation and is important in udder hygiene as repeated swings risk dirtier cups attaching: they are not washed between attempts, only between milkings. Accuracy also depends on teat placement (can be closer at end of lactation or if interval is too short), cleanliness of the udder, and presence of excessive hair.

With increased data from milk flow profiles during each teat’s milkings over time, the interval between milkings affects flow and therefore teat end pressures during harvest. Milk flow should begin shortly after cup attachment, reaching peak flow within 20 seconds, maintaining flow over the entire milking until harvest is complete. When flow reduces, this is detected by a flow sensor and the cup is released after a percentage flow (relative to peak) is reached. Delay in take-off can be customised if deemed appropriate. Abnormal milk profiles can be used to identify risks to teat health such as hyperkeratosis (see Figures 5-8).

Some machines deliver conventional left-right (L-R) pulsation and others can deliver circular pulsation. The advantage of circular is in the consistency of vacuum over the longer B-phase of harvest. Pulsation tube pressures set to, e.g., 43kPa will vary around 0.6kPa over the pulse in a circular system, whereas L-R pulsation may vary up to 1.6kPa. A more consistent vacuum results in a 63 per cent more consistent pulse and assists in a quick and complete harvest.

Post-treatment is mostly by spray mechanisms or by a nozzle within the cluster before take-off. The volume and velocity of ejected teat spray gives more effective coverage to produce a drip at the teat orifice. Accuracy in any system (including conventional) cannot be expected to be 100 per cent but should be monitored regularly to ensure teats are well covered.

Number of milkings and milk yield

Many AMS manufacturers aim for the number of milkings per day to be balanced against the herd yield and other 0 factors. In grazing systems, this changes over the lactation curve and season and can be targeted to 2.2-2.5/day at peak to around 1-1.5 at end of lactation. If changing from 2X milkings to more/day, an increase in yield has been reported as long as nutrition inputs and management allow. It has been published previously that up to 6X per day milkings can give more milk but the sample population in this group was nine cows (Dahl G.E., 2004). Figure 9 (page 485) shows 100,000 datapoints from eight Lely AMS farms, indicating no significant rise in milk yield beyond 4X. As flow profile data shows more slow starts and ends with less than six hours between milkings, milk access should be monitored to reduce risks of overmilking and detrimental teat health.

Mastitis risk and the AMS

There are many publications studying both bulk tank somatic cell count levels (BTSCC) and intramammary infections (IMI) in AMS technology and conventional systems but currently there is no outstanding evidence that one is significantly better than the other. A thorough review from 2018 suggests that BTSCC may rise over the transition from a conventional to an automatic system, as with any expected change in management, but returns to background levels, or less, after a short time (Castro A., 2018). As mastitis risk factors often involve barn and cow hygiene as well as equipment hygiene, attention should be on established mastitis control plans (Dohmen W., 2010). One of the greatest advantages of an AMS lies in early detection of mastitis. Milk can be monitored from every quarter on most milkings and the stockperson can be notified immediately when abnormal signals are found. As with conventional systems, if good hygiene and protocols are adhered to, the level of subclinical and machine-related pathogens become less frequent. Often barn and resilience-related pathogens such as coliforms are consequently brought to the forefront (Burvenich C., 2003). Risk reduction measures remain as with conventional systems: concentrate on hygiene and servicing; aim for good quality pre-treatment washes and teat sprays that provide disinfection, barrier, and emollient; and ensure that settings are correct for accurate delivery.

What is in the future?

The explosion of data from many of the robotic systems, whether milking, feeding or in manure management, has resulted in a potentially overwhelming information stream to the producer. Already, artificial intelligence and algorithms can read much of this and provide time-saving analysis and predictions based on real-time data from the equipment. Many aspects of milk harvesting can be customised: brush stimulation times, vacuums according to milk flow, and take-off, being only three examples. AI tools can view milk flow profile patterns over time and recommend optimum intervals between milkings and yields per milking to avoid teat end damage risk. There is much opportunity to increase this in the future.

Summary

Automatic milking systems are an increasingly mainstream part of dairy production, offering far more than labour-saving milkings. Successful AMS performance depends on the interaction between cow, barn, machine, and manager. Robots can milk consistently and gently, while collecting detailed data on milk flow, udder health, fertility, and efficiency. However, the technology only performs well when the basics are right: effective teat preparation, accurate attachment, post-milking teat treatment, appropriate milking intervals, and regular servicing. Barn design is equally important, with cow flow, lighting, water access, mobility, and grazing infrastructure all influencing voluntary visits. Robotic feeding and floor-cleaning can further improve consistency and cow comfort. Importantly, an AMS changes the farmer’s role from manual milking to proactive observation, maintenance, and interpretation of data. The future is likely to involve more AI-supported decision-making and increasingly individualised cow management.