14 June 1996

Tractor without a driver – not

so implausible

By Andrew Faulkner

DRIVERLESS tractors may seem the stuff of fantasies. However, the reality is nearer than you might think.

Silsoe Research Institute is more than two years into a project which, although majoring on auto plant/weed identification for spot treatment of inputs, has also developed a driverless concept vehicle to put that innovative husbandry theory into practice.

The Institutes Nick Tillett says his long-term vision – probably 10-15 years off – is of a lightweight tractor which could automatically self-pilot across a field, spot applying pesticides and fertiliser on an individual plant basis. Thats one step on from the current buzz technique of yield mapping and spatial variation.

Silsoe reckons up to a 90% reduction in pesticide use is achievable from per plant rather than conventional per area application.

Concentrating on the vehicle itself, the next step in the development chain is to create a commercial, as opposed to concept machine capable of standing up to on-farm operation – and abuse. Getting an engineer to tinker with computer settings is one thing, a layman quite another.

So how does the Silsoe auto pilot work? GPS, remote control, a piece of string? In fact, the basis of the system is no more than a standard monochrome TV camera, similar to those already used for security surveillance around farm buildings.

When travelling through the crop, this tractor-mounted camera sends visual images of the crop rows in digital format to a computer at a rate of 10 times/sec. This visual image is backed up by data, from a compass and wheel speed sensors, transmitted at 50 times/sec. By looking ahead, the vehicle knows both where it is and where it should be going.

Using this info, the computer sends a steering signal via a controller to hydrostatic drive motors on the tractors front wheels – steering is by relative speeding up/slowing down of the two front wheels, skidster-style to keep the tractor within the crop rows.

On the headland the camera switches off, and the compass and wheel speed sensors take over with a pre-programmed turning cycle to line up the tractor for the next bout.

Impressed? Well, it gets even cleverer. That same locating camera also produces what amounts to a continuous plant map as the tractor passes across the field. This map is sent in digital form to a second computer which is pre-programmed with row width information. In simple terms, the computer knows where the plants should be.

When the plant is not in a row, the computer assumes it is a weed and, if the tractor is fitted with a sprayer, can spot treat it.

Alternatively, the same system can be used to spot treat the crop rather than the weed – with foliar feed, for example. Hence, the term "plant scale husbandry": the identification and treatment of individual plants.

All this is some way off commercial application, as Silsoe readily admits. The system has to be more robust, and researchers also plan to build in more sophisticated plant/weed identification using differences in leaf texture and brightness.

There are probably more immediate practical uses for the guidance system than the vehicle itself. The tractor-mounted, sugar beet steerage hoe is just one machine which might benefit from using the Silsoe system for auto guidance. Some Eastern Counties posteriors, numb from hours perched on iron-clad hoe seats, would probably agree. &#42

What, no driver? Silsoe Research Institutes self-piloting concept vehicle is powered by an 8hp Honda petrol engine through a layshaft to hydrostatic units on the two front wheels. Under the tractors belly are a series of jetting nozzles for spot spraying treatment of individual plants.

Headland turning cycle has to be pre-programmed into one of the vehicles on-board computers.