5 thoughts on “Where is the difficulty of automobile autonomous driving?”
Kate
临 At present, autonomous driving is facing two great challenges, one is that the big data is not complete, and the other is that machines are in great difficulty to understand “human intention”. To solve the problem accurately and effectively, autonomous driving needs to ensure the completeness of the collected big data as much as possible. But the fact is that the data loopholes we now have hundreds of. For example, in 2015, there were 6.3 million traffic accident records in the United States, causing 35,000 people to die and 2.44 million injuries. But the actual number of traffic accidents may be 2-10 times that of official data. Before Google’s autonomous driving, CTO CTO Chris Urmson said in a speech that during Google, he had experienced a total of more than 200 kilometers of road tests, including 25 traffic accidents, but most of them were not available Caiting the police’s attention. “This situation is mostly small accidents. Autonomous cars make reasonable judgments, brake, and then humans hit.” Eumson said that many accidents often happened like this. Bleak
高 The high -precision map technology of automobile autonomous driving can be certain that the detection technology such as radar is up to one or two hundred meters more accurate, and no matter how far away, it is not expected. At this time, the high -precision map is very important. For example, the Audi A8 launched recently, while being equipped with Level 3’s autonomous driving technology, also has a new high -precision map system. It can not only update the driving conditions, calculate the route in real time, but also display the 3D simulation street scene map to provide more efficient efficiency Navigation services. According to estimates, the map accuracy of the complete autonomous driving phase requires 10 to 20 cm accuracy. The demand for autonomous driving technology for map updates must be second -level. Only vehicles can be equivalent to humans and even transcending human judgment time. Therefore, the update of map resources will be upgraded by cloud technology online (OTA). Of course, this is of course this. Of course It will make higher requirements for the Internet. In addition, the actual logo and the information of the storage logo may be different; under different lighting and weather conditions, various infrastructure logos are difficult to recognize. This requires the deep learning ability of the vehicle, but the relevant indicators may not meet the traditional perception algorithm. Bleak
It is certain that the radar technology will be one of the first difficulties for autonomous driving. As the basis of developing semi -autonomous driving and even autonomous driving, now traditional cameras and millimeter wavelera reach maturity. Lidar technology. The biggest problem of laser radar is that the cost is high. It is foreseeable that the solid -state laser radar will be the key development direction. Refining the resolution of product stability and smaller angle is difficult. Bleak
： Difficulties in navigation: how can the positioning be accurate at any place and time, and the GPS positioning will have no GPS signal at all in the underground garage due to the drift or loss of the building’s blocking signal; Drift or loss, or cannot be located because the feature points cannot be found; IMU, mileage, calculation positioning will accumulate errors; when multiple positioning means, how to design the positioning. Bleak
我们Net -optical technology is also the networking technology we often say, and the fashion is that V2V (car network), V2X (car union “Wanwu”) technology. There are many benefits of this technology. For example, it can track the real -time road conditions of each car in real time, coordinate the driving of each car, and arrange the route reasonably to avoid congestion and traffic accidents. Essence However, V2X technology is still only in the test stage. For example, the first self -driving test completed by the first Mercedes -Benz self -driving bus in Amsterdam can be automatically stopped or detoured at the bus stop, red light, obstacles and other locations automatically. It can be automatically realized at each bus station, and the tunnel will not affect the effect of autonomous driving. The reason for achieving this function is because they complete the communication between bus and base station. As for the V2V technology, there are already relatively initial applications. For example, many car’s multimedia system has the functions of the circle of friends, which can automatically search for nearby brand cars, and perform a communication function similar to WeChat, but this is limited to entertainment. Communication has not achieved the role of driving communication. [For example, Dongfeng AX4 just released before, its multimedia system has a communication function similar to WeChat. Bleak
临 At present, autonomous driving is facing two great challenges, one is that the big data is not complete, and the other is that machines are in great difficulty to understand “human intention”. To solve the problem accurately and effectively, autonomous driving needs to ensure the completeness of the collected big data as much as possible. But the fact is that the data loopholes we now have hundreds of. For example, in 2015, there were 6.3 million traffic accident records in the United States, causing 35,000 people to die and 2.44 million injuries. But the actual number of traffic accidents may be 2-10 times that of official data. Before Google’s autonomous driving, CTO CTO Chris Urmson said in a speech that during Google, he had experienced a total of more than 200 kilometers of road tests, including 25 traffic accidents, but most of them were not available Caiting the police’s attention. “This situation is mostly small accidents. Autonomous cars make reasonable judgments, brake, and then humans hit.” Eumson said that many accidents often happened like this. Bleak
高 The high -precision map technology of automobile autonomous driving can be certain that the detection technology such as radar is up to one or two hundred meters more accurate, and no matter how far away, it is not expected. At this time, the high -precision map is very important. For example, the Audi A8 launched recently, while being equipped with Level 3’s autonomous driving technology, also has a new high -precision map system. It can not only update the driving conditions, calculate the route in real time, but also display the 3D simulation street scene map to provide more efficient efficiency Navigation services. According to estimates, the map accuracy of the complete autonomous driving phase requires 10 to 20 cm accuracy. The demand for autonomous driving technology for map updates must be second -level. Only vehicles can be equivalent to humans and even transcending human judgment time. Therefore, the update of map resources will be upgraded by cloud technology online (OTA). Of course, this is of course this. Of course It will make higher requirements for the Internet. In addition, the actual logo and the information of the storage logo may be different; under different lighting and weather conditions, various infrastructure logos are difficult to recognize. This requires the deep learning ability of the vehicle, but the relevant indicators may not meet the traditional perception algorithm. Bleak
It is certain that the radar technology will be one of the first difficulties for autonomous driving. As the basis of developing semi -autonomous driving and even autonomous driving, now traditional cameras and millimeter wavelera reach maturity. Lidar technology. The biggest problem of laser radar is that the cost is high. It is foreseeable that the solid -state laser radar will be the key development direction. Refining the resolution of product stability and smaller angle is difficult. Bleak
： Difficulties in navigation: how can the positioning be accurate at any place and time, and the GPS positioning will have no GPS signal at all in the underground garage due to the drift or loss of the building’s blocking signal; Drift or loss, or cannot be located because the feature points cannot be found; IMU, mileage, calculation positioning will accumulate errors; when multiple positioning means, how to design the positioning. Bleak
我们Net -optical technology is also the networking technology we often say, and the fashion is that V2V (car network), V2X (car union “Wanwu”) technology. There are many benefits of this technology. For example, it can track the real -time road conditions of each car in real time, coordinate the driving of each car, and arrange the route reasonably to avoid congestion and traffic accidents. Essence However, V2X technology is still only in the test stage. For example, the first self -driving test completed by the first Mercedes -Benz self -driving bus in Amsterdam can be automatically stopped or detoured at the bus stop, red light, obstacles and other locations automatically. It can be automatically realized at each bus station, and the tunnel will not affect the effect of autonomous driving. The reason for achieving this function is because they complete the communication between bus and base station. As for the V2V technology, there are already relatively initial applications. For example, many car’s multimedia system has the functions of the circle of friends, which can automatically search for nearby brand cars, and perform a communication function similar to WeChat, but this is limited to entertainment. Communication has not achieved the role of driving communication. [For example, Dongfeng AX4 just released before, its multimedia system has a communication function similar to WeChat. Bleak