How do autonomous (driverless) vehicles perform when operating amid tough wintry weather conditions?
At the VTT Technical Research Centre located in Finland, they have developed Martti, a driverless vehicle that is capable of handling the challenges of snow and ice-filled roads. The vehicle is based on Volkswagen’s Touareg and it has had some success operating in the snow in the Lapland area of Finland traveling at a rate of 25 miles per hour. Carl Wellington, an engineer with Uber’s autonomous vehicle development team, says the snow presents an “interesting problem” for driverless vehicles. Up until this point, the majority of the early vehicle testing has been conducted in regions where there is no snowfall.
Driverless vehicles often use some variation of Light Detection & Ranging (LIDAR) technology that is based on the use of lasers to navigate. The lasers work in conjunction with radar to create a 360-degree scope where distances are calculated and the system interprets signs, signals, and roadway markings. Snowy conditions can reduce the functionality of this technology. The snow is often misinterpreted by the lasers as being other objects.
Matti Kutila, a project manager at VTT, says the Martti vehicle is designed to function better in snowy weather. The system is based on algorithms that function as adaptive filters that enhance the system of radar, lasers, cameras, and sensors. Kutila explains that Martti is not yet capable of operating properly on streets because the current street maps lack true accuracy. The technology in its current stage is likely to be implemented to transport skiers in the mountains and by the cruise industry to transport travelers through the terminal. VTT is among just a few organizations that have had some success with adapting to operation in the snow.
Carl Wellington says that the new sensors are capable of generating a good three-dimensional image of the environment in good weather conditions, but that the sensors are extremely sensitive to snowflakes. Uber has been conducting their snow testing in Pittsburgh and creating an algorithm that can decipher between snow and other objects. Chris Gerdes, a mechanical engineer with the Center for Automotive Research at Stanford, says that weather conditions present challenges in how the technology interprets the surroundings. The sensors can perform poorly in certain weather, which reduces the vehicle’s movement capabilities and lessens friction with the road, potentially causing the vehicle to spin.
Karl Iagnemma, CEO of nuTonomy, says they are testing cars in the snow in Boston now and future plans include also doing so in Detroit. He explains that the three-dimensional view that self-driving vehicles interpret is altered with the presence of accumulations of snow, creating difficulties with operation. He points out driverless vehicles have a potential advantage over humans when operating in snow. This is based on the ability of the technology to detect lane marks and crosswalks regardless of whether they are coated with ice or snow. It seems that further testing and slight changes could begin to improve the operation in wintry conditions very soon.
Laszlo Kishonti, of a start-up company called Imotive, recently began testing in Finland during the winter, after conducting some testing in lighter snow closer to their headquarters in Hungary. He says that the vehicle must be capable of enhanced awareness to safely drive in snow, as sensors tend to become blocked. He explained that the vehicles need the capability to adapt based on the conditions, as a human might. He feels that reducing the speed of operation amid snow is one critical part.
Chris Gerdes says that some automated vehicles are really imperfect for operating in good weather with limitations in extreme conditions. Many manufacturers have determined the weather conditions that are best suited for their vehicles and now must adapt to others. The industry is increasingly seeking out locations with extreme weather conditions for putting their vehicles to the test. Many vehicles are operating fine is rain, yet are not capable of handling a real blizzard type of environment. He used an analogy regarding teenage drivers; stating that parents choose to let teens drive while knowing that it is dangerous—but they will improve over time.
As it is, the smart car, self-driving vehicle industry keeps moving forward, inventing ways to make use safer on the roads in the near future.
