Unmanned technology is maturing so that fans of the ecstatic meal. According to a Reuters report on February 10, U.S. regulators recognized Google’s driverless vehicle as a driver, a decision widely considered to be an unmanned milestone.
However, if you are encouraged by this, start thinking of getting on your car and wearing a seat belt and giving instructions to get to the good life of your destination, only to be disappointed in the short term. Even if the test mileage has accumulated up to 100 million miles of Google driverless car, want to successfully pass the exam to get a driver’s license, I am afraid have to check the weather forecast, pick a good weather.
Because, snow days, driverless car this new “driver” will strike!
Blizzard makes driverless aerial vehicles “blind”.
The development of unmanned locomotive technology can be described as ever-changing, but now it has encountered the challenge of nature: snow days. According to the February 12 report by Bloomberg in the United States, Volvo’s driverless car test recently encountered snow and snow weather test.
Marcus Rothoff, director of Volvo Car’s driverless car program, said: “It’s really hard to overcome such a bad weather, but we can overcome it with ice on the pavement, but once we get off the snow, especially in front of the car The driverless car sensor is likely to fail. ”
In fact, in tackling the stormy weather on the road, many car companies go on going. According to Google’s unmanned vehicles report released in December last year, Google unmanned vehicles are also testing the impact of bad weather, but in the event of heavy rains, unmanned vehicles will still be automatically parked on the roadside waiting for traffic improvement. Ford Motor Co., which began researching driverless technology since 2005, started the unmanned ice and snow weather test in January of this year. Ryan Eustice, an associate professor at the University of Michigan who works for Ford, points out that driverless technology is “almost resolved” by the media and the public with exaggerated propaganda and understanding, and a driverless car that can adapt to all kinds of road, weather and conditions Still the ideal goal.
Why do snowy days blindfolded “eyes”?
Even for the humans, snow and ice day driving a car is a challenge, and snow-capped days have more of an impact on driverless technology. “The reason is mainly the environmental perception difference between unmanned vehicles and human pilots.” Liu Yong said, an associate professor of robotics laboratory at Zhejiang University School of Control Science and Engineering who has spent many years in the field of robot environment perception. Liu Yong explained that human beings driving through the eyes to identify accessible areas and objects, and the impact of changes in the weather on human visual system and cognitive system less. Unlike humans, autonomous driving requires more than just the visual system. It also requires constructing a three-dimensional map to determine its own position and route it to its destination in real time.
Unmanned systems often contain multiple sensors, of which laser radar is widely used to construct three-dimensional maps and is a core element of unmanned systems. In the snow and ice weather, the reflection characteristics of the road caused by icing or snow changes cause the effect of lidar to be affected, thus affecting the construction of the 3D map and causing the “location dizzy” of the unmanned vehicle. In addition, in addition to sensing the three-dimensional environment, unmanned vehicles also need visual sensors to identify information such as lane lines and road signs in order to achieve obstacle avoidance and follow regulations. Ice and snow weather is also a challenge for road markings. For example, lane lines and road signs may be partially covered with heavy snowfalls, while houses on both sides of vehicles and roads may be more difficult to identify due to snow and ice cover.
Remove the” snowstorm.” on the way of driverless.
Liu Yong analysis, you want unmanned vehicles in the snow and ice can also smooth travel, you need to build a stable three-dimensional map and reliable road signs recognition considerations. For the former, it is expected that sensors that are more suitable for inclement weather such as rain and snow will be developed in hardware, and stable detection can be performed regardless of whether the road is covered with heavy snow or heavy rain. From the algorithmic level to the use of precision maps constructed in good weather, it is also important to complete smooth sensing and planning in the snow and ice in conjunction with existing maps. For the latter, Liu Yong believes that the problem of stable recognition under different weather and seasons is an important issue in computer vision. By constructing datasets under various weather and seasons and using deep learning algorithms that have achieved numerous achievements in recent years, It is possible to achieve reliable identification close to human beings.
It is understood that, in order to solve the camera, the problem of laser sensor vision, Google to the roof of the laser sensor with a “wiper” to ensure that the rain is not obstructing the field of vision, the sensor found that after the rain, it also had to teach the car can penetrate Over raindrops and cold condensate exhaust smoke to see things, continue to accurately detect objects.
“We wanted unmanned vehicles to drive most of the weather,” said Jim McBride, head of unmanned vehicle technology at Ford, so simulating the real world is so important: “UAVs still need to walk before technology updates It’s a long way to expose the existing technology to the environment, and the process is refined and learning to adapt. “