The Oracle Robot

The Oracle robot is a Sheep Sheering robot and it was developed at the University Of Western Australia in 1979. Many of the Oracle's concepts formed the basis for its successor, the Shear Magic Robot. The Oracle is pictured in Fig.1.

During shearing operations, the sheep is restrained firmly by holding its legs and its head in a manipulator, called the ARAMP. The shearing arm is directed by complex motion control algorithms which maintain the cutter at a predefined height above the sheep's skin. It should be noted that not all sheep are the same shape, and they sometimes object very strongly to being shorn.

The robotics arm which 'holds' the clippers can be manoeuvred in six directions, and is powered by a series of hydraulic actuators using proportional analogue servo controls. These actuators are controlled by a minicomputer through a conventional D/A interface. The Hewlett Packard 21MX-E is utilised to control Oracle.

In order to maintain such a critical distance from the sheeps' skin, many Sensors are used to feedback the actual position of the shearing arm and the relative position of the sheeps' skin. Difficulties are associated with such sensing, as wool has a tendency to conduct electricity, and this characteristic can vary with moisture, humidity and other environmental factors, so sensors can give erroneous readings. However, a combination of several types of sensors are used to keep the clippers at a safe distance from the skin whilst still ensuring a reliable cut to the wool.

The Oracle robot uses a software model of the surface of a sheep in order to have a preliminary 'idea' of the task ahead, and to determine an approach route to the sheep. When the robot is within a predetermined distance from the sheep, the sensor inputs take over. The cutter follows a surface path computed from weight and dimension data for the sheep to be shorn. This path is then modified by the minicomputer depending on data received from the various sensors. The surface model is predicted on the basis of statistical correlation with a large 'sheep database'. Further to this concept, the Oracle can 'learn' from each sheep shorn, by using the actual surface encountered and the predicted surface to obtain an error, which is then taken into account for the next animal.