Agriculture - a challenge to robotic science

Citation

Marchant, J. (1998), "Agriculture - a challenge to robotic science", Industrial Robot, Vol. 25 No. 5. https://doi.org/10.1108/ir.1998.04925eaa.002

Agriculture - a challenge to robotic science

Agriculture ­ a challenge to robotic science

The authorJohn Marchant is Guest Academic Editor of this issue of Industrial Robot and is based at Silsoe Research Institute, Bedford, UK.

Keywords Agriculture, Robots

In a perfect robot-friendly world cauliflowers would grow in exactly straight lines, cows would stand completely still, oranges would all be on the outside of a leafless tree, the sun would shine from a constant direction and some would say pigs might fly. The real agricultural world is, however, very different. In contrast to an industrial situation where, for example, each component on a production line should be the same, variability is ever present in agriculture. Despite the best efforts of plant and animal breeders, agricultural products even if genetically identical are quite different when measured in engineering terms.

The requirement to deal with the natural world means that agricultural robotics is difficult. Few, if any, robotic systems have reached commercial realisation. However, some are close, for example, milking, mushroom harvesting and grass cutting (the latter being an amenity rather than an agricultural application). Thus, agricultural robotics is mainly the preserve of the research scientist and engineer. Natural variability generally means that any agricultural robot needs to sense changes in the products to be handled. Perhaps the most investigated sensing technique is machine vision where the attractions include the non-contact nature of the sensor, the large amount of data delivered, the cheap and commonly available hardware, and the realisation that humans and other animals make very effective use of vision. The downside to machine vision is that it is at least as difficult as robotics, especially when dealing with natural situations.

Vision is a significant component in the recent work on mushroom harvesting at Silsoe Research Institute (Plate 1) where the algorithms use the reasonably well defined shape and surface reflectance characteristics of the mushrooms. One way of coping with the difficulties of agricultural robotics is to try to control the environment or the objects to be manipulated. For example, for many years plant breeders and horticulturalists have produced apple trees whose shape and size make harvesting easier for humans. This was taken one stage further in French work where grape vines were produced to allow easier pruning with sensor-controlled machinery. It may be possible to control difficulties caused by natural lighting. For example, in the Agribot orange harvester the natural light is effectively swamped by using a laser pointer to target the fruits. A further technique is used in this work where the part that a robot finds difficult ­ locating the fruit ­ is done by a human, whereas grasping and removing the fruit is done by the robot.

Plate 1 Robotic mushroom harvester

If agricultural robots are to succeed, they must be cheap compared with their industrial counterparts. In agriculture, profit margins are small and equipment is often used seasonally rather than constantly as on a production line. This means that agricultural robotics will look to mass market applications to supply this basic equipment. Helpful factors here include cheap and powerful computing for the office and home market, sensors and information systems from the automotive area and visual hardware from the home communications and entertainment areas. Agricultural robots may also need to be more robust but may only require modest accuracy. A good example of getting the balance right is the work on robotic milking at Silsoe Research Institute now nearing commercial reality (Plate 2). Situations where robots interact with animals are one of the few areas where the manipulated objects can fight back. Cows can (and do) kick the robot and the robot could harm the cow if it wasn't gentle in its action. A direct drive pneumatic system has been used to power the axes, the air providing a cushion against damage to either robot or workpiece.

In such a short article it is impossible to do justice to the volume of creative work being done by researchers around the world. In summary, the agricultural setting provides a significant challenge for robotics. It will be some time before we see as many robots in fields as there are on the production line, but a start has been made with some robotic devices being on the verge of commercialisation.

Plate 2 Automatic milking system