In regions with expensive labour or a shortage of labour, automatic milking is serious alternative for a traditional milking parlour.
A successful use of automatic milking depends largely on the farm conditions and the management skills of the farmer. '
The number of farms milking with an automatic milking system has increased rapidly since 1998.
The first ideas about fully automating the milking process were generated in the mid seventies. Cost of labour in several countries was growing and this was one of the main reasons to start the development of automation around milking. An important step was the development of reliable cow identification systems. The first applications were automatic concentrate feeders. A further step in the automation of milking parlours was the development of automatic cluster removers. In the early eighties, automation in milking parlours was expanded with the development of milk yield recording equipment and sensors to detect udder health problems. All these developments and new milking technology reduces the labour input during milking, resulting in a higher output per man-hour. In many parlours, the task of the milker was limited to udder preparation, teat cup attachment and control of the cow and the milk. The final step in the automation development seemed to be the development of automatic teat cup attachment systems. The idea of course was to develop a fully automated automatic milking system (AM-system).
Rossing et al (1985) concluded it was possible to milk the cows in a concentrate feeder. A concentrate feeder was used to build an one cow milking parlour. Cows could enter the concentrate feeder 24 hours a day. When cows entered, the milking cluster was attached manually. The development of automatic cluster attachment was a real challenging step. At the end of the eighties and the beginning of the nineties, a series of cluster attachment principles were in study in several research institutes all over Europe. However, it took almost a decade to convert the techniques for locating teats and attaching teat cups to fully integrated and reliable automatic milking systems. The first milking robots were installed on commercial dairy farms in the Netherlands in 1992. The breakthrough of automatic milking came at the end of the nineties. At the end of 2001, over 1100 farms world-wide milked their cows automatically. In Europe, almost all milking machine companies have an AM-system in their range of products and automatic milking has become a fact instead of fiction for many farmers and their families. This paper gives an overview of the principles of AM, the developments in Europe, the chances and challenges and the impact of AM on the management of a dairy farm.
Farms with Automatic Milking Systems
The first AM-systems on commercial farms were implemented in North Western Europe. The reasons for these countries to start the development of AM-systems most probably were related to the expensive labour in these countries and the farm structure with family farms. Increasing costs of labour, land and buildings and machinery, while milk prices tended to decrease, forced farmers to increase their output per man-hour. Average herd-size showed a continuous increase and this phenomenon is still ongoing. In some countries, such as The Netherlands, the tax system makes investments more interesting.
Research institutes performed the first studies in The Netherlands, but co-operation with industrial partners came very fast. The first AM-systems on commercial dairy farms appeared also in the Netherlands. This is probably because the first industrial suppliers of AM-systems were Dutch. Subsequently, several other companies started also to develop AM-systems (DeLaval, Insentec, Orion, Westfalia-Surge). Other milking machine companies used existing teat cup attachment technology and combined that with their own milking technology to develop an AM-system (Fullwood, Gascoigne-Melotte, Bou-matic, Manus).
Figure 1. Number of farms with automatic milking systems (source PV-Lelystad)
In the first years after the introduction of the first AM-systems, the adoption went slow, until 1998 (Figure 1). From that year on, in the Netherlands automatic milking became an accepted technology by a large part of the dairy sector and in the same period also other countries adopted AM-systems, like Germany, Denmark and France. At the end of 2001, worldwide more than 1100 commercial farms used one or more AM-systems to milk their cows. More than 90 % of all dairy farms with an AM-system are located in northwestern Europe. Most dairy farms with an AM-system can be found in the Netherlands. However, AM-systems are also regarded to have potential for as well the USA (Reinemann and Jackson-Smith, 2000) as Canada (Rodenburg and Kelton, 2001).
Principles of Automatic Milking Systems
An AM-system has to take over the “eyes and hands” of the milker and therefore these systems have electronic cow identification, cleaning and milking devices and computer controlled sensors to detect abnormalities of the AM-system, cow and milk. An AM-system consists of six main modules:
- Milking stall
- Teat detection system
- Robotic arm device for attaching the teat cups
- Teat cleaning system
- Control system including sensors and software
- Milking machine (including system cleaning)
1. Milking Stall
AM-systems can be divided into single stall systems and multi-stall systems. Single stall systems have an integrated robotic and milking system, while multi-stall systems have a transportable robot device. Each stall has its own milking devices, like in a milking parlour. The global capacity of single stall and multi-stall systems is presented in table 1. The design of the milking stall is still based on the concentrate-feeding box. Automatic milking relies on the cow’s motivation to visit the AM-system more or less voluntarily. The main motive for a cow to visit the AM-system is the supply of concentrate; therefore, all AM-systems are equipped with concentrate dispensers.
Table 1. Global capacity in number of milkings per day for AM-systems (source: AM-suppliers)
||Maximum herd size (cows)
||Number of milkings per day
|One single stall
|Two stalls (multi-stall)
|Three stalls (multi-stall)
|Four stalls (multi-stall)
2. Teat Detection System
The udder shape and teat position will differ substantially from cow to cow and is dependent on facts as lactation stage, amount of milk and milk interval and deformation of the udder shape due to lying in the cubicle. Moreover, cows, although they are locked up, can move in the milking stall and so the position of the teats will change. Udder and teat position can be measured and stored in a database, but it is impossible to attach teat cups successfully based on only this information. An AM-system therefore needs an active teat detection system to localise the teats. Using different techniques, like ultrasonic systems, laser techniques and CCD camera systems, has solved this technically quite difficult process (Artmann, 1997). All these techniques are used to find the position and place of the teats in reference to a fixed point at the robot arm. In fact, the system creates a three-dimensional view, so the system knows where to attach the teat cup to the teat. The environment where these techniques have to operate is quite rough; moisture, dust and manure and may influence the teat detection system. Therefore special attention has to be paid to these circumstances, otherwise the performance of the system will decline.
3. Robotic Arm
Different types of robot arms are used in automatic milking systems (Artmann, 1997, Rossing et al., 1997). Some robot arms imitate conventional milking by using an arm with a gripper, which picks up the teat cup from a storage rack at the side of the stall. The four teat cups are attached in succession. Besides teat cup attachment, some robot arms keep the teat cups in the correct position, and also carry out teat disinfection after milking.
Multi-stall systems have a movable robotic arm. Milking device and robot arm are separate units. Each milking stall has 4 milk cups mounted on a special rack. The robot arm moves from one stall to another, picks up the whole rack or each separate teat cup and starts the attachment process. After attachment, the robot arm is disconnected and can be moved to the next stall. Another system (one stall system) uses a robot arm with a fixed milk rack integrated with the robot arm. The robot arm is fixed to the milk stall, so each stall has it’s own robot arm.
For all systems, each teat cup is attached separately, starting with the teat cups for the hind teats and finishing with the front teats. In general the attachment process is quite time consuming, somewhere between 45 and 100 seconds, depending on the behaviour and the udder characteristics of the cow and depending on the AM-system. A skilled milker needs 10 to 15 seconds to attach all four teat cups.
4. Teat Cleaning System
It is common practice to clean the udder and teats before milking. Also milk ejection has to be stimulated and udder health has to be checked. The purpose of cleaning teats is primarily to remove dirt and other particles that can contaminate the milk. Cleaning is also necessary to meet (inter)national legislation and hygiene rules from the dairy industry. Research and practical experiences at conventional milking systems suggest that if udders and teats are clean, teat washing can be suspended without affecting milk quality. However, current AM-systems do not have sensors to detect the amount of dirt on the teats, so the cleaning system is always based on average dirty teats. Efficient cleaning is of particular importance with high levels of spores in the environment of the cow. The teat cleaning system should also aim at minimising the risk of transferring udder pathogens from teat to teat or from cow to cow. To prevent residues in milk, teats should be free from disinfectants before attaching the teat cup.
There are four principles of teat cleaning with AM-systems.
- Sequential cleaning by brushes or rollers
- Simultaneous cleaning by a horizontal rotating brush
- Cleaning with water in the same teat-cup as used for milking
- Cleaning in a separate ‘teat cup like’ device.
Generally, plain water is used for teat cleaning. Sometimes the cleaning devices are flushed with plain water or disinfected between consecutive cleanings. The efficiency of the current cleaning devices to clean the teats appears to be a matter of concern, although little information is available (Lind et al, 2000). Initial trials with automatic cleaning devices showed that cleaning with these devices is better than no cleaning, but not as good as manual cleaning by the herdsman in conventional milking systems (Schuiling et al, 1992).
Besides cleaning the teats, automatic cleaning devices also stimulate the milk let down process. Stimulation of the milk ejection reflex is necessary for efficient milking. Little information is available about the differences between the different teat cleaning methods with respect to the intensity of the milk ejection reflex, but it is clear that the way of pre-treatment in AM-systems before each milking is very repeatable compared with conventional milking. This might have a positive effect on the milk ejection. Bruckmaier et al (2001) showed that teat cleaning devices in AM-systems are suitable for pre-stimulation and they concluded that AM-systems can fulfil the physiological requirements of dairy cows to induce milk ejection as a prerequisite for complete milk removal.
5. Control System and Sensors
AM-systems need sensors to observe and to control the milking process like the milker is doing. Therefore, AM-systems are equipped with a variety of sensors. These sensors should be the ‘eyes’ of the AM-system and their task is to control the milking process and to detect any abnormalities. Each AM-system is equipped with sensors to control the technical functioning of the AM-system, like cow identification, teat cup attachment, vacuum level, start of the milk let down process etceteras. Most AM-systems have also other sensors to control the quality of the milk process, e.g. check on abnormalities in the milk, milk yield, electrical conductivity (udder health), temperature of the milk, feed intake, body weight and so on.
All measurements are automatically stored in the database and a management program is used to analyse these data and to control the settings and conditions for the cows to be milked. Attention lists and reports are presented to the farmer by screen or printer messages. In urgent cases like a break down, or severe problems with a cow, the system immediately warns the herdsman by sending a text message via a mobile phone or another alarm.
6. The Milking Machine
The milking machine is more or less similar to the milking techniques in conventional milking parlours, except that there is no milking cluster. This also means that individual quarter milking is applied. In many systems also individual teat cup detachment is applied. For each quarter a teat cup, milk and pulse tube is used and the milk will be transported separately till the milk meter or the milk receiver. In general milk tubes will be much longer than those applied in conventional milking, resulting in a considerable vacuum drop below the teat during milking. An air inlet below the teat cup is used to transport the milk from teat cup to receiver. Due to the length of the milk tube, the air/milk ratio in AM-systems is almost 10:1 compared with 3:1 in conventional milking. Most AM-systems have a modified boiling water cleaning or a circulation cleaning as system cleaning and special unit- and cluster flushing systems (Schuiling et al., 2001)
Automatic Milking and Management Aspects
Changing over from a milking parlour to automatic milking will lead to big changes for both herdsman and cow and can cause stress to both. With automatic milking, the milking process does not require permanent supervision anymore. However, this does not mean that the hours spend on traditional milking will be spared. New labour tasks arise with the implementation of automatic milking in the farm; control and cleaning of the AM-system, twice or three times a day checking of attention lists including visual control of the cows and fetching cows that exceeded maximum milking intervals. Little field data is available on the labour savings when applying automatic milking. Several model studies showed physical labour savings of 30 to 40 % compared with conventional milking systems (Sonck, 1995, Schick et al., 2000). Ipema et al., (1998) and Van’t Land et al.,(2000) reported labour demands for AM-systems from 32 minutes up to 3 hours per day. On average a 10% reduction in total labour demand is reported compared with the conventional milking system with twice milkings per day (De Koning, 2001). The biggest change however, is the change in the character of the labour. Instead of mainly handwork during milking, the herdsman has to check several times per day attention lists from the computer of the AM-system. Decisions have to be taken accordingly. However, the work is less time bound compared with the milking parlour system, thus enabling a more flexible input of labour. This can be especially attractive on family farms. On the other hand, because milking has changed into a 24-hour process, system failures can occur for 24 hours per day. Therefore, there should always be a person on duty to react on system failures. Practical experiences show that system failures occur approximately once every two weeks. Good maintenance and attendance can decrease the number of failures. For instance, sensor failures might occur because the sensors are dirty, because they were not cleaned. This type of failures can easily be prevented by a well-organised working method.
The impact on the cows also, can be big. The AM-system might not be suitable for all cows, because of udder shape and teat position or behaviour. Nevertheless, the culling rate of cows, because they are not suitable for automatic milking, is estimated to be less than 5-10%. More important is the introduction period, cows should be handled quietly and consistent, to learn them to adapt themselves to the new surrounding and milking system. Automatic milking places emphasis on the cow’s motivation to visit the AM-system to be milked more or less voluntarily. For this reason all AM-systems are equipped with concentrate dispensers. In the transition from conventional to automatic milking, cows have to learn to visit the AM-system at other times than before. Special attention is needed and in the first weeks, human assistance will be necessary.
Figure 2. Frequency distribution of milking intervals in hours over a 2 year period
In practice, the number of milkings per day varies from 2.5 till 3.0, but rather big differences within the milking intervals are reported from commercial farms. De Koning & Ouweltjes (2000) found that almost 10% of the cows realised a milking frequency of 2 or lower over a two year period milking with an single stall AM-system (figure 2). This occurred even though cows with a too long interval were fetched three times per day. These cows will not show any increase in yield or even might show a production loss. By changing the milking parameters of the AM-system, it is quite easy to prevent cows from being milked at low yields or short intervals. But it is much more difficult to prevent cows from being milked with long intervals. This means it will be necessary to manage the intervals by fetching cows that have exceeded a maximum interval. Usually this is done several times per day at fixed times around the cleaning procedures of the AM-system. However, fetching cows cannot guarantee that long intervals are prevented as the data showed. Fetching cows three times per day that have exceeded an interval of 12 hours, means that the maximum interval may amount to 20 hours. Fetching cows with shorter intervals is quite time-consuming, however.
One of the main benefits of automatic milking is an increase in milk yield from more frequent milking. It is known that milk production, in terms of milk production per hour, is dependent on the milking interval. An increase in milk yield from 6 to 25% in complete lactations has been shown when increasing the milking frequency from two to three times per day (Erdman and Varner, 1995). Dairy herd information records in The Netherlands show that daily milk production increases by 11.4 %, when farms change from milking two times per day in a milking parlour to automatic milking (unpublished data). French data show an average 3 % increase in milk yield up to 9% increase for farms that utilized the AM-system for more than 2 years (Veysset et al., 2001, Billon, 2001).
Important aspects of successful implementation of an AM-system are the attitude and expectations of the dairy farmer (Hogeveen et al., 2001). When expectations are too high, disappointments will also be high. Automatic milking requires, especially at the start period, high input of labour and management. Almost all manufacturers of AM-systems have had customers who afterwards, went back to a traditional milking system. The exact reasons are not always known. Key factors of a successful implementation of AM-systems are:
- Realistic expectations
- Good management support by skilled consultants before, during and after implementation
- Flexibility and discipline to control the system and the cows
- Ability to work with computers
- Much attention to the barn layout and a good functioning cow traffic
- Technical functioning of the AM-system and regular maintenance
Cows should have easy access to the milking system, using the cows motivation for eating. The main motive for a cow to visit the AM-system is the supply of concentrate. AM-systems are equipped with concentrate dispensers, to attract cows to the AM-system to be milked. Therefore, the routing in the barn should be according the Eating – Lying – Milking principle. Cows should have an easy access to the milking stall. Selection gates, long alleys and so on should be minimised. The AM-system should be a part of the free stall barn (Lind et al., 2000). A central position of the AM-system in the barn minimises the walking distances of the cows. However, for matters of hygiene, in many countries the dairy industry requires the placement of the AM-system close to the milking room. Moreover, it is required that the AM-system can be reached through a clean route.
After visiting the milking system, the cow should have access to the feeding area. Using this milking-feeding-lying principle, the cows are motivated to use the AM-system. Moreover, throughout 24 hours, sufficient roughage should be available. This is a prerequisite to have optimal cow traffic. There does not seem to be a big difference in average milking frequency between the one way and the free cow traffic systems in practice (Ipema et al., 1998; Van’t Land et al., 2000). The one-way cow traffic system is a very effective way to utilize the AM-system and to learn for cows. However, there is a consensus that for animal welfare, free cow traffic is better. Cows spend more time in the waiting area in a one-way cow traffic system (Hogeveen et al., 1998). Especially for AM-systems with a high occupancy rate, this might affect the number of visits to the roughage station and thus result in a limited intake of roughage. A combination of one-way and free way system is also possible: only some cows are forced to go to the milking system first, others can pass through a special selection gate to the feeding area.
In most European countries, grazing during summer time is a routine or even compulsory. Moreover, from a ethological point of view, many consumers in North Western Europe believe grazing is essential for cows. In the Netherlands, a substantial proportion of the farms with an AM-system is applying grazing, showing that grazing in combination with AM is possible (Van ‘t Land et al., 2000, Ruis-Huitinck et al., 2001). One dairy industry in The Netherlands started with paying a bonus on the milk price when grazing is applied.
Capacity of an AM-system
The capacity of an automatic milking system is often expressed as the number of milkings per day, but this number will largely depend on the configuration of the automatic milking system, like the number of stalls and the use of selection gates, milking frequency, machine on time, herd size and cow traffic system. Increasing the number of milkings per cow per day, does not necessarily contribute to a high output capacity in kg of milk per day. This is due to the more or less fixed handling time of the automatic milking system per milking and the decreasing amount of milk per milking when cows are milked more frequently. Milk flow rate and yield have a large impact on capacity in kg per day (De Koning and Ouweltjes, 2000). By changing the milk criteria for individual cows, the AM-system can be optimised to realise a maximal capacity in kg per day. Besides effects on capacity, negative effects of certain milk intervals, such as increase of FFA with shorter milking intervals and possibly an increase in SCC with long and short milking intervals must also be taken into account.
Milk Quality and Cooling
Milk quality is without doubt one of the most important aspects of milk production on modern dairy farms. Milk payment systems are not only based on fat-content, but also on milk quality. Consumers aspect a high quality level of the milk products they buy. Although automatic milking uses more or less the same milking principles as conventional milking, there are some differences, as already explained.
Results from commercial farms indicate, that in many cases milk quality is negatively effected (Klungel et al., 2000; Justesen et al., 2000; Van der Vorst et al., 2000, Pomies et al., 2000, Billon, 2001) after the introduction of automatic milking. Results show almost a doubling of the bacterial counts, although the levels are still relatively low and far within the penalty levels. The cleaning of the milking equipment and the cooling of the milk seem to be the most important factors regarding the increase in bacterial counts. Also cell counts are not reduced after the change to automatic milking, despite the increased milking frequency. It is not clear if this phenomenon is related to the AM-system or to the changes in operational management. Special attention should be given to the housing conditions of the cows. The same studies showed also a significant increase in FFA levels. This increase cannot be explained solely by the shorter intervals, because the increase of FFA with AM-systems is bigger compared with conventional milking parlours milking three times per day. Another explanation may be the increased air inlets during teat cup attachment, milking and at take off. Also the cooling system may play a role.
Milk should be cooled within 3 hours to a temperature below 4 oC. The basic requirement is that the system can handle the specific conditions of automatic milking. In general there are four principles to adjust the cooling system to automatic milking; 1) indirect cooling with an ice-bank tank, 2) combination of bulk and buffer tank, 3) storage tank with modified cooling system and 4) instant cooling. For the ice bank tank and the modified cooling system, it may be useful to have an additional buffer, which is able to store the milk when the bulk tank is emptied and cleaned. This enables the AM-system to continue milking, thus increasing the capacity of the system.
The number of farms milking with an AM-system has increased rapidly since 1998. In regions with expensive labour or a shortage of labour, automatic milking is serious alternative for a traditional milking parlour. The introduction of automatic milking has a large impact on the farm and affects all aspects of dairy farming. Important aspects that require attention are management, barn layout, milk quality and grazing. A successful use of automatic milking depends largely on the farm conditions and the management skills of the farmer.
Billon P. (2001). Les robots de traite en France; impact sur la qualité du lait en le système de production; In : Proceedings : Il Robot di Mungitura in Lombardia ; Cremona, Italy
Erdman, R.A. and M. Varner (1995) Fixed yield responses to increased milking frequency. Journal of Dairy Science 78: 1199-1203
Ipema A., H.J. Schuiling (1992) Free fatty acids, influence of milking frequency, in: A.H. Ipema et al (editors) Proceedings of the Symposium on Prospects for Automatic Milking, EAAP publication 65, pp 244-252, Wageningen Pers, Wageningen, The Netherlands
Ipema A, A. Smits and C. Jagtenberg (1998) Praktijkervaringen met melkrobots, in: Praktijkonderzoek, 98-6, pp 37-39
Hogeveen, H., A.J.H. van Lent and C.J. Jagtenberg (1998). Free and one-way cow traffic in combination with automatic milking, in J.P. Chastain (editor) Proceedings of the Fourth International Dairy Housing Conference, ASAE 01-98, pp 80-87, Michigan, U.S.A.
Klungel, G.H., B.A.Slaghuis, H. Hogeveen (2000), The effect of the introduction of automatic milking on milk quality, Journal of Dairy Science, 83:1998-2003
Hogeveen H., Y. van der Vorst, C. de Koning, B. Slaghuis (2001) Concepts et implications de la traite automatisée, in: Symposium sure les bovines laitiers, pp 104-120, CRAAQ, Quebec, Canada
Koning C.J.A.M. de, and W. Ouweltjes (2000) Maximising the milking capacity of an automatic milking system, in: H. Hogeveen and A. Meijering (editors) Robotic Milking, pp 38-46, Wageningen Pers, Wageningen, The Netherlands
Koning, C. de, (2001) Automatic milking; chances and challenges: in: A.Rosati, S. Mihina, C. Mosconi (editors), Physiological and Technical Aspects of Machine Milking, pp 131-140, ICAR TS 7, Rome, Italy
Justesen, P. and M.D. Rasmussen (2000) Improvement of milk quality by the Danish AMS self-monitoring programme. Page 83-88 in: H. Hogeveen and A. Meijering (editors) Robotic milking. Wageningen Pers, Wageningen, The Netherlands.
Land, A. van ‘t et al (2000) Effects of husbandry systems on the efficiency and optimisation of robotic milking performance and management, in: H. Hogeveen and A. Meijering (editors) Robotic Milking, pp 167-176, Wageningen Pers, Wageningen, The Netherlands
Lind O.,A.H. Ipema, C.de Koning, T.T. Mottram, H.J. Herrmann (2000) Automatic milking, Bulletin of the IDF 348/2000, pp3-14, IDF, Bruxelles, Belgium
Pomies D., J. Bony (2001) Comparison of hygienic quality of milk collected with a milking robot vs. with a conventional milking parlour, in: H. Hogeveen and A. Meijering (editors) Robotic Milking, pp. 122-123, Wageningen Pers, Wageningen, The Netherlands
Reinemann, D.J. and D. Jackson Smith (2000) Evaluation of automatic milking systems for the United States. in: H. Hogeveen and A. Meijering (editors) Robotic milking, pp 232-238, Wageningen Pers, Wageningen, The Netherlands.
Rodenburg, J. and D.F. Kelton (2001) Automatic milking systems in North America: Issues and challenges unique to Ontario; in: National Mastitis Council Annual Meeting Proceedings, pp 163-169, NMC, Madison, WI, USA.
Rossing, W. and P.H. Hogewerf (1997) State of the art of automatic milking systems. Computers and Electronics in Agriculture 17: 1-17.
Rossing, W, A.H. Ipema and P.F. Veltman (1985). The feasibility of milking in a feeding box. IMAG Research Report 85-2, Wageningen, The Netherlands.
Ruis-Heutinck, L.F.M., H.J.C. van Dooren, A.J.H. van Lent, C.J. Jagtenberg and H. Hogeveen. (2001) Automatic milking in combination with grazing on dairy farms in the Netherlands (abstract). In proceedings of the 35th Congress of the International Society for Applied Ethology, Davis, U.S.A.
Schick, M., M.-R. Volet and R. Kaufmann (2000) Modelling of time requirements and milking capacity in automatic milking systems with one or two milking stalls. Page 32-37 in: H. Hogeveen and A. Meijering (editors) Robotic milking. Wageningen Pers, Wageningen
Schuiling H.J. (1992) Teat cleaning and stimulation, in: A.H. Ipema et al (editors) Proceedings of the international symposium on prospects for automatic milking, EAAP publication 65, pp 164-168, Wageningen Pers, Wageningen, The Netherlands
Schuiling H.J., J. Verstappen-Boerekamp, K. Knappstein, C. Benfalk (2001), Optimal cleaning of equipment for automatic milking, Deliverable D16, http://www.automaticmilking.nl/
Sonck, B.R. and J.H.W. Donkers (1995) The milking capacity of a milking robot. Journal of Agricultural Engineering Research 62: 25-38
Veysset P., P. Wallet, E.Prognard (2001) Automatic milking systems: Characterising the farms equipped with AMS, impact and economic simulations, in: A.Rosati, S. Mihina, C. Mosconi (editors), Physiological and Technical Aspects of Machine Milking, pp 141-150, ICAR TS 7, Rome, Italy
Vorst, Y. van der, and H. Hogeveen (2000) Automatic milking systems and milk quality in the Netherlands, in: H. Hogeveen and A. Meijering (editors) Robotic Milking, pp 73-82, Wageningen Pers, Wageningen, The Netherlands