Now, more and more new cars are equipped with L2 level automatic driving assistance system, so that the car can "drive itself", so that the driver's hands and feet can be relaxed in long-distance driving, and the vehicle can be operated without nervous tension at any time. But sometimes it may be "too easy", accidentally handing over the safety of driving to the machine, and not knowing it!
Automakers also have a variety of names for driver assistance systems, such as Ford's Co-Pilot360 and Toyota's Safey Sense, which are common driver assistance systems on the market. Tesla is more confident, directly named Autopilot assist, does it sound higher and bigger? But no matter how the name changes, the word "assist" will definitely appear on the official website, publicity and even in the manual, in order to tell the owner that the driving assistance is only an auxiliary after all, and when the system still has a misjudgment, the owner should still pay attention to the road conditions at any time, and if necessary, even need to intervene artificially, and take the driving safety into his own hands.
Humans look at the road through their eyes, but what about cars?
Before exploring the assistive system, let's first think about how we "look at the road" when we drive. Is it through the eyes to see as far as possible to observe the road ahead; coupled with the rearview mirror assist, and then through the brain based on driving experience to make a judgment, determine whether the car should accelerate, slow down or turn. The translation into machine language is: information gathering - decision-making - performing actions, which is easy for humans with driving experience, but it is doubly difficult for a computer that intentionally makes one or says one thing or another.
The eyes of the car: radar, camera
Cars usually use radar and cameras to replace human eyes to understand the road conditions, and driver assistance systems are often matched with a variety of sensors. For example, ACC cruise control typically uses medium- and long-range radar to scan the road ahead and collect information about metal obstacles (other vehicles) in front of it. AEB Active Brake integrates the ACC radar with the image from the forward camera lens for interpretation and database comparison. These various sensors are equivalent to the "eyes" of the driver assistance system, responsible for the front-end action of data collection. With sufficient information, it can be handed over to the computer to judge and make decisions according to human beings, and then use electronic throttle signals, electronic brakes, electronic steering assists and other control components to perform the precise and complex action of "driving".
Although the speed of computer execution is fast today, the driving assistance system has long been popularized, but because the road conditions are ever-changing, the computer can not be like humans with "experience" to make "prejudgments", so the driving assistance system will often be slow and half a beat, there is always some time difference with the current road conditions, as for how to make the computer drive like an old driver linear and smooth, that is where the skills of the car factories are tuned!
The computer will also be unable to see clearly, and of course it will make mistakes
Before entering the next stage of explanation, we first distinguish what is the difference between the above left and right pictures, I believe that as long as the driving with normal vision can see that the left side is the empty highway in front of us, and the right side is a simple white picture. But in the eyes of the driver assistance system, the two pictures may be "completely indistinguishable"! This is because the performance of our eyes, whether in terms of brightness and shade tolerance or picture resolution, is basically far more than any camera, and the computational efficiency of the brain is far better than any computer on the market, and many of the light and shadow changes or road conditions that we are accustomed to may not be able to do anything for the driver assistance system.
White is the enemy of the self-driving system
For example, in 2016, Tesla's first autopilot caused by the death of the car accident, is because the camera can not distinguish between the "white container car" and "white sky" across the lane, resulting in the vehicle without slowing down, rushing under the truck, and similar cases are still emerging, enough to illustrate that the driving assistance system with the photographic lens as one of the main judgments, for the color is too close to the object / background, or complex light and shadow transformation is unable to distinguish, then if the driver relies too much on the driver assistance system, Accidents can easily arise.
Radar cannot detect stationary objects
In addition to the possibility of mistakes in the photographic lens, the medium and long-range radar carried by the ACC fixed speed cruise is also a mistake-prone device, because the ACC radar cannot distinguish the stationary obstacles in front, is it a vehicle? Or a large signpost? Or a roadside substation box? Therefore, most ACC ignore the stationary objects at the beginning of the design to avoid frequent misjudgment, thereby slowing down and braking sharply causing greater safety hazards. Only when the car in front of it gradually slows down, the ACC radar can correctly and accurately detect and react, and once the speed difference is too large, the system can easily identify it as a stationary object and crash directly into it.
Secondly, due to the characteristics of radar, it mainly detects the reflection of objects, thereby calculating the direction, height, size and speed of the object in front. Therefore, if it is an obstacle to an abnormal vehicle, such as a platform trailer, a locomotive, a pedestrian, etc., it is likely that the software database "does not know" and causes "blindness" and cannot issue the correct instructions to slow down the vehicle.
Humans and driver assistance should be 1+1 greater than 2
Driving this thing, humans and machines should be complementary to the existence, humans have excellent eyes and brains, can quickly correspond to various road conditions, but the disadvantage is easy to fatigue and distraction, driving assistance has a lot of restrictions, specific circumstances will not be seen, but the machine will not be tired and not easy to be affected by external factors, the two can really improve driving safety!
Imagine driving an L2-level system on a fast road backlit at dusk, with a malfunctioning car in front of us, parked in the middle of the lane without lights on. According to the above-mentioned camera, radar restrictions, at this time the camera may be too large because of the light contrast and can not see clearly, ACC radar also ignores the stationary objects in front of it and continues to accelerate, the only thing that can avoid the accident is the driver who continues to pay attention to the road conditions and sits in the driver's seat! Seeing this, do you still dare to drive an L2-level intelligent driving car while playing with your mobile phone?