Published on TECH.EU
Link to publication
When you think of autonomous cars, your mind probably wanders to some sunny neighbourhood in Silicon Valley with a Google car cruising around in the middle of its latest test. A snowy and icy street in Russia is perhaps not what you had in mind.
In the race for autonomous and connected vehicles, Russia’s Cognitive Technologies is making its own play for the market that everyone from Apple to Volkswagen wants in on.
Cognitive Technologies has been around the block. It was founded in 1993 as a software firm, building image and character recognition software for the enterprise. Along the way it sold its voice recognition solutions to Intel and its text parsing tech to Yandex.
“Three or four years ago, our co-owners decided to expand the business,” explained Roman Tarasov, vice president of global business. Now the company has something else in mind – autonomous cars, and specifically an autopilot system that detects objects and obstacles ahead of the vehicle.
In 2014, it announced this foray into the burgeoning and still-rocky world of autonomous and connected vehicles. Engineers at Cognitive began developing several advanced algorithms that could aid vehicles and drivers (for now anyway) in avoiding collisions and fatal accidents.
C-Pilot, according to Yuri Minkin, Cognitive’s head of self-driving department, is modelled on the hippocampus, an important component of the human brain that assists in spatial navigation. It determines the most important information at a given moment so that the car can react aptly.
“The system identifies everything which is near the car, 360 degrees, and it identifies the horizon line and the cars, pedestrians, motorcyclists, bicycles. It uses enhanced artificial intelligence to predict what each object is going to do,” added Tarasov. “For example, the system sees a kid and an old lady. The kid is much more likely to jump in front of a car than the old lady. The system analyses the scene and predicts possible behaviour of the objects and acts accordingly. That’s our killer feature.”
Cognitive refers this as “foveal computer vision”, focusing on the centre of the field of vision and as objects enter the frame, like a car entering a lane or a pedestrian stepping off the footpath.
C-Pilot uses less computing power, added Tarasov, as it only stores what it deems to be the most important information that it has gathered.
We’re marching further towards a world a where cars drive themselves but we face several obstacles in the way; the obvious being safety. Tesla grabbed headlines in July when a driver was killed in a car crash while using the car’s autopilot feature. The controversy led to serious questions around the competency of the feature and if Tesla had over-promised on its capabilities.
Tarasov explained that Cognitive has built the system for what he calls “real roads”, where conditions aren’t the best, road markings are unclear, or road signs are obscured or missing. And in the case of many places, the weather is bad, like the snow-covered roads of Russia: “As we’re a Russian company in Russia we have specially designed the system for Russian roads.”
This is one of the key differentiators from other systems. “It’s better than basically any system on the market because it’s able to work in real world conditions,” he said.
It’s a big claim. There’s plenty of competition in the space. The obvious example is Israel’s Mobileye, which is a former partner of Tesla’s but is now working with Delphi Automotive and Intel and will be staging a demonstration for its car next month at CES in Las Vegas.
Cognitive has designed its own driverless car, refurbished from a Nissan X-Trail, which it has been using to carry out tests on autonomous driving and to put its image recognition solution through the rigour.
“We’re now in the process of negotiating with major car manufacturers,” said Tarasov as the company is determined to bring this technology to markets outside of Russia, which it can’t do on its own.
Currently its biggest partner is Russian truck manufacturer Kamaz. The pair has a deal in place that will see the truck maker implement C-Pilot in a line of trucks in 2017 to assist drivers. Ultimately Kamaz plans to manufacture self-driving trucks by 2020.
Cognitive has a strong base in its own country, which is all well and good, but not conducive to reaching global markets.
“In Russia we’re a very [well known] company. The western markets, the global markets are not very much aware of us but we can use our good contacts within our Russian car manufacturers to get to the western ones. That’s the plan,” admitted Tarasov.
“Basically we need western investors to get to the car makers.”
Cognitive hasn’t raised any VC funding at all. It has received between $8-10 million from the Russian ministry of education and science. Seeking funding, ideally from a company in the automotive sector, may be on the horizon if Cognitive Technologies wants to make people think of Russia when chatting about autonomous cars.