Cognitive Technologies Claims World First AV Harvester

A Russian driverless vehicle company is claiming the first market ready autonomous combined harvester.

Cognitive Technologies’ ADS Agrodroid retrofit system was put through its paces in front of international media at trials in Siberia. It was clear from the trials that the driver doesn’t touch the steering wheel although he’s still in the cabin doing another job, controlling parameters of harvesting from cutting to threshing, cleaning and unloading grain. Before an autopilot had been installed, distraction between the tasks, magnified by the drivers’ tendency to exceed the optimum speed, resulted in up to 18% losses of yield, says Viktor Karbyshev, CEO at Mezheninovsky. The self-driving unit helps the human to concentrate on just one task. It also helps to reap the harvest in a shorter time working around the clock with three shifts of human drivers, a sound advantage in Siberia’s short summer.

The AI-employing self-driving system used in this harvester, unlike GPS-based autopilots from other developers, is not dependent on telematics or OTA connectivity, making the tech usable in distant areas with poor infrastructure.

Regulations and insurance

Insurance of autonomous agricultural equipment remains an issue. “Insurers have not yet sorted out a policy towards autonomous ag machines,” says Olga Uskova, founder and CEO at Cognitive Technologies. “For the pilots, we’d elaborated a palliative scheme of certifying our system as an OEM-approved modification. It satisfied them.”

It is unclear how to classify and regulate self-driving farm machinery. Industry-specific AV standards and certification is an empty space, says she, but it will soon become urgently needed: “Farmers will want to know how to leverage stronger and weaker solutions.” She suggested that some log of the frequency of enforced human interventions could be applied. The, of course, yet another set of regulations is required to allow autonomous machinery into the public motorways.

For the pilot project, regulation was both an enabling and a limiting factor. On one hand, safety rules for farmlands are much less rigorous than that for public roads or manufacturing and commercial facilities. On the other hand, Agrodroid could be used more extensively if a legislative framework allowed but at the demonstration in Tomsk region of Siberia it functioned as an intelligent cruise control assistant when in the field, steering the harvester along an edge of unmown field and avoiding obstacles.

Harvest of data

Low price-sensitivity of clients in agriculture compared to that in the automotive sector is another reason behind the early start of commercial use. The Agrodroid system costs about $6,000 per unit. “As soon as we were able to drop the price below 3% of the total cost of a vehicle, it’d become comfortable for customers,” says Uskova. The manager argues that, from a point of view of a single AV developer, the agricultural market can be more lucrative than the automotive one despite smaller fleets thanks to after sales servicing.

“User support is what promises to make the tech lucrative,” she explains. Long-term co-operation with customers is also important for further improvement of the technology. Each day of harvesting brings more region-specific visual datasets necessary to train the convolutional neural network, Agrodroid’s key component, says Andrey Geltser, head of C-Pilot department in Tomsk at Cognitive Technologies.

Farmers’ uptake

Commercial exploitation has started this year with eight harvesters in Siberia and another 800 units pre-ordered for the next five years. Uskova also says that a major global manufacturer of agricultural machinery is studying Agrodroid with prospects of signing a supply contract with the developer.

“My father said that it’s easier for us to send a man to space than make a good tractor,” says Karbyshev. “We have long been waiting for this system and clearly understand benefits of high tech.”


Leave a comment

Your email address will not be published. Required fields are marked *