Agricultural autosteer is no longer considered science fiction.
In the last two years, autosteer technology based on global positioning systems has become reliable and affordable and can be found on many Canadian farm implements. It’s available from different companies and comes as a factory option or a kit for on-farm installation.
A team of ag engineers at the University of Illinois recently developed an autosteer system that is more field-based than GPS-based. It uses digitalized information from a stereo vision camera mounted on the front of the tractor rather than depending solely on GPS satellite signals or a local tower for differential correction.
Read Also
VIDEO: Green Lightning and Nytro Ag win sustainability innovation award
Nytro Ag Corp and Green Lightning recieved an innovation award at Ag in Motion 2025 for the Green Lightning Nitrogen Machine, which converts atmospheric nitrogen into a plant-usable form.
A triangle is the basis of any geometric locating system because a minimum of three specific points of reference are needed to determine location. The more reference points available, the closer positions can be pinpointed.
Once a precise location has been determined, the data can be used to steer a tractor, sprayer, combine, mini-van, fishing boat or all-terrain vehicle.
The Illinois engineers conducted successful video camera autosteer experiments for a number of years, but wanted to make their John Deere 7700 autosteer tractor more self-reliant and responsive to real field conditions. Their aim was to make the self-steering tractor less dependent on reference data from outer space or a tower.
To accomplish this, they needed two locating reference points on the tractor that worked in conjunction with the distant target image on which the camera focuses. This would give them the three reference points needed for accurate field-based autosteer.
Their first video-steered tractor used a camera with only one lens. They used the tractor-mounted camera as one reference point, along with any stationary object the camera picked up in front of the driverless tractor as it drove up and down the field. The camera would automatically pick its own target.
However, that provided only two points of reference and a two-dimensional picture. They still required a third point of reference from GPS to achieve depth of perception.
The latest system, which they call 20/20 Vision, uses a stereo camera with two lenses, giving them the necessary three points of reference on the ground and producing digital three-dimensional pictures.
The 3D camera they selected is a Videre Divine PH-MD1 from California.
Ag engineer Qin Zhang said the old 2D cameras lacked the third geometric point and also had trouble with changing light levels. The 2D cameras would lose contrast and thus the ability to distinguish between crop rows and background soil. He said the 3D camera solved that problem.
The tractor had already been modified with an off-the-shelf electro-hydraulic steering valve, which was directed by the data controller in what is now commonly known as a conventional autosteer instalment.
Zhang said electro-hydraulic parts are basically the same. The only thing that changes from one system to another is the information flowing into the data controller.
“We developed our own software system to integrate the image from the stereo camera into the data controller, which drives the hydraulic steering. The new software takes the two images from the camera and references them as two distinct points on a triangle.”
The engineers have logged 100 hours on the 20/20 stereo vision tractor.
“It’s pretty flat terrain, so we imposed curved rows along with straight rows to give it a real good test,” Zhang said.
The highest working speed recorded for scientific purposes was 13 km-h, at which it stayed on track. The engineers did get it up to 23 km-h, but had failures at that speed.
“The frame we built for mounting the camera to the tractor wasn’t strong enough, so the camera started bouncing. When the camera bounces, it gives us incorrect data.”
Zhang said a different camera mount will solve the problem, perhaps something like the Steadycam used by news videographers to take the shake out of the picture. He is confident they will achieve total accuracy at speeds up to 19 km-h.
The new software package, called sensor fusion, brings together in-field data gathered by the stereo vision camera and the conventional GPS data. GPS gives an exact starting location in the field while the stereo camera provides data about the tractor’s motion as it relates to the path it must follow. Sensor fusion then sends messages to the data controller that steers the tractor.
The engineers have also incorporated an important safety factor. When the camera senses that the tractor is approaching a stationary object, it alerts the sensor fusion software. Sensor fusion then commands the data controller to stop the tractor.
“Depth perception is what allows the stereo camera to alert the data controller if it approaches an object,” he said.
“We couldn’t do that before with a single lens camera. We had to depend on GPS signals for our triangle.”
He said they still use GPS to locate the tractor in the field, but now the guidance triangle is on the ground in front of them.
“That’s much more accurate, I think.”
He said the 20/20 technology is ready for market.
“Everything you need for autosteer is already factory installed on modern tractors. The only thing you need for commercialization is the sensor fusion software and the stereo camera.”
Zhang doubts that totally self-guided implements will become common but he does think there’s potential for the 20/20 technology to be incorporated into existing autosteer implements.
He said the operator will be the first to benefit from the system.
“We’re hoping to make farming more enjoyable by reducing tedious and intense work levels.”
Another possibility is for big farms. He thinks a lead implement, such as a tractor or combine, could be operated by a qualified driver. This lead implement could be followed through large fields by a fleet of cheaper, self-guided implements taking their primary instructions from the lead implement.
