ÐÓ°ÉÔ­´´

If you thought the British drink a lot of milk (5.5 billion litres last
year), just consider the amount of time spent producing it. There are 2.7
million dairy cows in Britain, spread among 30 000 farmers – an average
of 90 per herd. One person can milk and tend up to 150 cows, each producing
an average of 6000 litres of milk annually. Each cow is milked twice a day,
and it takes the human attendant about a minute to prepare the cow and attach
the automatic machinery to the four teats on the animal’s udder. The extraction
process, based on techniques first used in 1860 and little changed since
flexible rubber teat cups were introduced in 1906, takes another 5 or 6
minutes.

That means that people spend a total of 90 000 hours each day just attaching
cows to machines. Certainly it provides employment, but the demands of
efficiency make milking hurried and tedious, in a noisy, dirty environment.
There is a steady turnover of skilled farmhands because they catch diseases
from the cows and become disabled because of the repeated strains on the
back and legs. Lack of time may also mean that the more stimulating and
rewarding work outside the milking parlour, such as calving cows or training
calves to suckle, is neglected.

The generally accepted practice of twice-daily milking is, in any case,
a compromise between a reasonable working day for the humans and the cows’
continuous milk secretion. The only way to milk more often without placing
impossible demands on people is to automate the process completely so cows
can be milked, preferably at their own instigation, without continuous human
attention. In short, a robot milkmaid.

A fully automatic milking machine that cows could attend when they wanted
would be a boon to farmers, and potentially a huge earner for its inventor.
Compared to the cost of labour, it could be made profitable even for a herd
of 60 cows. It is also important that the animals themselves benefit: cows
that are milked more frequently produce more milk and are less susceptible
to infection of the udder (mastitis). John Webster, professor of animal
husbandry at the University of Bristol, also suggests that a system which
cows attended voluntarily ‘would be better suited to the physiological needs
and passing whims of the dairy cow’.

The elements of milking are quite simple. In Britain, for hygiene reasons,
cows must now be milked in a clean area. They are gathered and held in a
collecting yard so that the operator of the milking parlour can give them
a few seconds of individual attention. Before milking, the teats must be
cleaned and inspected for cuts and infections. Then the teat cups are attached:
the milking machine draws out milk by a combination of suction and the pulsation
of its flexible liner over the animal’s teat. When the milk stops flowing,
the cups are removed and the teats sprayed with lanolin and disinfectant.

For the attendant, the flow of work is continuous; but cows may stand
waiting for up to two hours before and after each milking. Even so, half
of the time in dairying is spent milking the cows and cleaning up afterwards.

CONCENTRATED RESEARCH

For several years commercial companies and research institutes in Europe
have been trying to automate the whole milking process. Much of the research
has concentrated on developing methods of locating the cow’s teats and attaching
the cups using a sensor-guided robot arm.

But the difficulty for anyone designing a milking robot is that most
industrial robots work on inanimate, rigid objects that have a defined shape,
orientation and velocity; they usually blindly follow preprogrammed paths,
grasp something well defined and move it to a predetermined location. However,
animals are soft, and move and change their shape unpredictably. An effective
robot needs to be ‘soft’ too, but softness is not in great demand among
industrially made robots. Spring-operated mechanisms on or near the gripping
end offer one solution, while air-driven systems let the whole robot yield
under excess forces.

Attaching the teat cups is also a demanding business. A cow is about
2 metres long, 1 metre wide and 1.5 metres high. The area of the udder is
about 0.64 square metres, while the teats are about 10 centimetres long
and up to 25 centimetres apart. That would present no problem, except the
cow keeps moving and so do the teats. To be effective, robots must be guided
by sensors, yield in a collision and be able to track normal movements caused
by the animal’s breathing and general shuffling about. They do not need
much preprogramming: given sensor guidance, a feedback loop corrects positional
errors and makes precision irrelevant.

A new milking robot developed in Holland by a company called Prolion,
and demonstrated at a trade fair in Amsterdam last January, uses ultrasound
to locate the udder and then the individual teats. However, the product
is not for sale yet. Another technique being developed jointly by a team
from the University of Maryland in America and Gascoigne-Melotte, also a
Dutch company, identifies the cow from a transponder on its body and guides
the robot after accessing a computer database detailing where the teats
are located on each cow’s udder.

But a cow is always liable to kick any mechanism to pieces, emphasising
the fact that any device intended for agriculture must be cheap, reliable
– and robust. Some researchers have tried to solve the problem of how to
manoeuvre and attach teat cups by restraining the cow. Though this does
not injure the animals, they may feel trapped and become unwilling to turn
up for future milkings.

With funding from the Agriculture and Food Research Council, the Silsoe
Research Institute based near Bedford has developed a robot which uses pressurised
air to control rapid response to movements. The prototype takes a maximum
of 100 milliseconds to react to the cow’s movement. Pressurised air is
a very good medium for driving a low-cost system using simple cylinders
and pistons, but makes exact positioning more difficult than for hydraulic
or electrical systems. However, the components of the drive are cheaper;
as a milking robot will inevitably suffer collisions and kicks, softness
minimises damage to both robot and cow, while careful design ensures that
sensors are only an irritation to the cow. It is made of aluminium and stainless
steel so that the system can be washed; the electronic systems are tightly
sealed to withstand high-pressure cleaning needed for hygienic reasons.

ENCOURAGING STEPS

In the Silsoe prototype, passive methods (such as the shape of the stall,
with food at the front) encourage the cow to adopt a milking posture: with
her front feet on a step 200 millimetres deep, her teats swing clear of
her hind legs. Four T-shaped pneumatic ‘paddles’ (one on each side at the
front and rump of the cow) push gently against the cow to sense her position
and movements in the milking stall. Many cows now wear a transponder that
identifies them, so the positions of the teats on the udder can be stored
and retrieved using a database. A robot arm with a positional accuracy of
+-2 millimetres holds the cup, and when that is within 20 millimetres of
a teat’s estimated position, a local sensor comprising a matrix of light
beams provides correction data – based on which beams are blocked off by
the teat – to guide the cup directly beneath it. If the teat enters the
cup, the suction of the milking machine ensures that it becomes attached.
When all four cups are on, the robot withdraws. As attachment now takes
just over 1 minute while milking takes 4 to 7 minutes, future robots could
service more than one stall.

In a trial last November, the Silsoe system milked 10 cows three times
a day for 12 days and achieved an 85 per cent attachment rate. Some industry
sources reckon 90 per cent would be good enough for a commercial product.

Work has also begun on automatic techniques to inspect the udders for
lesions or dirt, and if necessary to clean them or divert the cow to a
‘sickbay’ for human attention. Software is being developed so the robot
can identify and recover from situations such as the cow standing on the
milking vacuum pipe. But mechanising the complete process requires a knowledge
both of engineering and of how cows behave.

HAPPY COWS

Experience so far suggests that cows have little fear of robots. In
fact, they enter the milking stall more readily – probably for food – than
they leave it. Mark Varner, a researcher on the Gascoigne-Melotte project,
comments ‘They may just like to be milked.’ But do they? Equally, frequent
milking increases tissue development in the udder: can this make the milk-providing
period (300 days every year) longer, so that cows do not have to have a
calf every year to produce milk? And will cows be ‘happier’ if they decide
when to be milked or fed?

Another benefit of automatic milking is its possible use in monitoring
the health and welfare of cows. A lot of information about individual dairy
cows is already gathered manually by the farmer or farmhand and then fed
into a computer to determine an animal’s health, production efficiency
and breeding status. It is essential in herd management to identify the
short period when the cow is fertile. Monitoring milk yield and appetite
helps determine the cow’s health, while unusual changes in the milk’s conductivity
can indicate mastitis, while its temperature helps indicate changes in body
temperature. Automatic milking systems will inevitably provide more data
on individual animals, so processing of sensor information will enable the
early identification and treatment of many subclinical ailments. The result
should be better animal health, higher milk quality – and of course more
time for herders to do other tasks.

Farmers and milking machine engineers say that there will be a ready
market for this technology when it becomes reliable. But it may be up to
five years before any is viable. However, with potentially enormous sales
worldwide, a system costing a few tens of thousands of pounds would be a
small addition to a typical milking parlour which already costs between
£80 000 and £90 000 to equip. Previous revolutions in agriculture
have begun with engineering developments – think of the steel plough and
the tractor. The robot may well initiate another.

Toby Mottram, who works at the Silsoe Research Institute, is a former
cowman who retrained as an engineer. Mike Street, also an engineer, set
up the project.