Tesla Autopilot in the eyes of a longtime user

Lei Feng network: According to the author of this article Xia Hao, Xiaopeng car founder and CEO.

A murder case took place in the United States soon after, and a former U.S. Army poster soldier named Joshua Brown was killed in the road. What is puzzling is that this seemingly ordinary blood case has caught the attention of the world. What is the reason? Let's go to the crime scene to find out.

Incident time: May 7, 2016 afternoon

Location of the incident: High-speed road section in Florida, USA

Involved: Joshua Brown, 18-wheeler truck, and it - Tesla's Autopilot system

This is the world's first automatic driving accident death case. This accident directly pushed the currently hotly debated autopilot technology to the cusp of public opinion.

Many people can not help asking, Tesla is not claimed to be able to automatically drive it? Should this accident be the responsibility of Tesla Autopilot? In this regard, the National Highway Traffic Safety Administration (NHTSA), the highest regulatory body of US automobile safety, is still investigating the working conditions of the Tesla Autopilot system at the time of the incident, but it is certain that the driver It is responsible for its non-monitoring behavior.

As a long-term user of the Tesla Autopilot system, the author has had a personal experience for each version since Tesla launched the autopilot function. Regarding Tesla’s automatic driving system safety, I would like to talk about my own point of view.

| Autopilot Tesla exemption tips

1) Tesla's manual. Each autopilot assist feature on the Tesla user manual has an eye-catching red exclamation mark warning alert. E.g:

About Lane Assist: Lane Assist only provides guidance and is not a substitute for direct visual observation. It is the driver's responsibility to remain vigilant, pay attention to the lane of the drive, and keep an eye on other road users. Failure to do so can result in serious injury or death.

About Active Cruise Control: Active cruise control is designed for driving comfort and convenience and is not a collision warning or avoidance system. It is the responsibility of the driver to stay vigilant, to drive safely, and to control the vehicle. Be sure to observe the road ahead and be prepared to take corrective action at any time. Failure to do so can result in serious injury or death.

With regard to collision avoidance aids: Pre-crash warnings only provide guidance and cannot replace focused driving and accurate judgment. Be sure to observe the road conditions while driving, and do not rely on front collision warnings to provide early warning of possible collisions. Relying on a forward collision warning to warn of a potential collision may result in serious personal injury or death.

2) Tesla's meter. Tesla provided automatic driver assistance technology instead of drone technology, and Tesla still needed manual monitoring. This can also be confirmed by the reminder displayed after the Model S turns on the automatic driving mode: “Please always hold the steering wheel and be ready to take over. If the system finds that your hands are not on the steering wheel, it will even warn you and gradually Reduce the speed until your hands are back on the steering wheel again."

3) Tesla is reminding the driver to take over all the time. When Tesla started automatic driving, even if the driving conditions were good, it will automatically jump out of the interface after about 15 minutes, and the buzzer sounded, please hold the steering wheel. Otherwise, the vehicle will slow down and stop. This measure is also aimed at preventing the danger brought by the driver's laxity.

Tesla's automatic driving function reminds the user to hold the steering wheel

| The case is repeated to restore the truth

Let us restore the accident scene at the time based on information disclosed by Tesla and the news media.

On May 7th, the sunny Florida weather was warm and a Model S was driving on a flat highway. The dashboard shows that Autopilot is currently available. Owner Joshua Brown launched his favorite autopilot function and gave control to the car. The vehicle is in good condition and everything is normal. Inside the vehicle, multiple sensors are periodically acquiring information to perceive the surrounding driving environment. The intelligent control system makes decisions and controls the vehicle according to the environmental parameters. The millimeter-wave radar installed in the front of the vehicle monitors the position and speed of the front object. The Mobileye Q3 camera on the windshield also looks directly in front of it to determine the surrounding vehicles and obstacles. At this moment, a white body-trailer near the front intersection is crossing the road. Tesla is only 100 meters away. The sunshine is too strong, and the white body makes it difficult to tell whether the front is a white cloud or a car.

"Hey, is it a car? It doesn't seem to move. The location is also a bit high. It looks like it's not a car. Hey, camera, what do you think?" asked the radar.

"Is there a car in front of me? Well, I haven't seen it yet. It's very empty in front of it. The white clouds are blossoming." The camera replied.

"Then it should be preceded by a large signboard. I say how high the position is. Uh, there is no vehicle in front of you. You can pass it. Let's move on."

In this way, it fooled Tesla's Autopilot system.

A few seconds later, the tragedy happened. Model S drilled into the bottom of the trailer, and the front windshield collided with the bottom of the trailer. The driver passed away on the spot.

Since the accident was still on the vehicle, the above result was restored based on the information recorded by the Tesla pickup vehicle. Tesla’s official statement: “Under strong sunlight, neither the driver nor the automatic driver noticed the white color of the trailer. The body, therefore, failed to start the brake system in time. As the trailer is crossing the road and the body is high, this special situation causes the Model S to collide with the bottom of the trailer when it passes the bottom of the trailer.”

Therefore, in the case of Tesla's statement, we can believe that the above process is the actual accident. To sum up, the causes of the accident are two aspects: one is the failure of the perception module carried on Tesla, and the other is the dereliction of duty of the pilot's master.

The failure of the perception module mainly includes two aspects:

First, millimeter-wave radar mistakenly detects the body of the trailer as a sign. According to Elon Musk's description, the millimeter-wave radar measures a huge obstacle in front of it. Due to the excessive reflection area of ​​the trailer and the high ground clearance body, it is more like a suspension from a millimeter-wave radar point of view. Traffic sign above the road.

Excessive body and excessive reflection area are understandable, but the detected position is suspended above the road in question. The main purpose of millimeter-wave radar is to detect information on obstacles on the road surface. It is not information on traffic signs that are suspended above roads. It is difficult to determine exactly what the front object is by simply using millimeter-wave radar, and only obstacles can be determined. Shape, size and position. The height of the trailer certainly does not have a high traffic signage. Therefore, when the millimeter-wave radar detects an obstacle and judges it as a traffic sign, it is really puzzling.

Incident site environment and accident

Second, the visual perception module suffers from failure. Tesla is equipped with the Mobileye EyeQ3 camera sensor, and some people guess that because of the blue sky and white clouds at that time, the camera's visual system judged the white body as white clouds, which is only speculation. We can confirm that according to Tesla, the Mobileye EyeQ3 has misjudged the white body passing through the road and failed to recognize the obstacle.

According to the above two comprehensive judgments, many coincidences came together to cause this accident. The deeper technical reason may be that the fusion algorithm of the millimeter-wave radar and the visual perception module has insufficient algorithm verification, or there is a bug in the algorithm.

From the driver’s point of view, Tesla’s automatic driving originally required the driver to be vigilant and ready to take over. Tesla is also constantly reminding users of their manuals, disclaimers, and driving. However, according to relevant reports, Brown is likely to be watching DVD-played "Harry Potter" movies in the car instead of observing the traffic.

If this is true (and only Mr. Brown, who has already passed away, knows what he was doing at the time), the main cause of the accident can basically be attributed to Mr. Brown’s reliance on Tesla’s autopilot system to control dereliction of duty.

Tesla's Autopilot system does not detect trucks with technical problems, but Tesla has repeatedly warned drivers to pay close attention to the road conditions and is ready to take over the steering wheel. Excessive reliance on and overuse of this feature has made this potential risk hidden. Even if it only has a probability of one ten thousandth, it will be at the expense of living life.

| Status of Automatic Driving Technology

NHTSA divides driverless technology into five phases:

Level 0 is that the driver has 100% control and the vehicle does not have any safety system auxiliary equipment;

Level 1 is that the vehicle owns single or multiple independent functional electronic control systems, such as an automatic emergency braking system;

Level 2 is at least two controls that can be automated, such as combining an active distance control cruise control system with a lane maintenance system;

Level 3 is the vehicle has automatic dodge barriers, self-guidance, active control and other functions, but the driver still has control;

Level 4 is fully automated driving of the vehicle. The user only needs to give related information, such as destination, route, etc., which means that the driver can engage in other activities in the car, such as working on the Internet, entertainment or rest.

In the industry's view, the development of autonomous driving technology is a gradual process, and its ultimate stage will develop to unmanned driving. Right now, the so-called automatic driving of major car companies is actually adding more intelligent assistive systems so that cars can complete certain specific “autopilot” actions. From a rigorous conceptual point of view, the contents and boundaries of autonomous driving and driverless coverage are significantly different. According to this classification, the current global autopilot technology still does not reach the level of unmanned stage 5, that is, Level 4 has full automatic driving capability.

Google first tested its self-driving cars in Austin, outside California last year, before it was only tested on public roads in Silicon Valley. This year, the test sites increased again, Kirkland, Washington, Phoenix, and Arizona also became test sites. The vehicles used for testing are all equipped with specially trained drivers and are not yet open to the public. Google’s patents in recent years include those concerning the switching of driverless status, the reminder of passengers, and gestures to control cars. We can see that Google’s autopilot technology is being tested in more road environments, and at the same time it is conducting a number of researches on autopilot services. However, it still fails to reach the Level 4 level of commercial civilian use, and it is only monitored by people. On the premise of unmanned driving to advance driverless technology services.

In addition to Google, the current technology level of most other OEMs is around Level 2. Tesla is also in the transition phase from Level 2 to Level 3. In other words, as Tesla advertises, the autopilot technology it provides requires manual monitoring. The main driver of driving is the human. The so-called “Autopilot System” provides more auxiliary functions.

| Correct posture for autopilot - always vigilant

It cannot be denied that the powerful features of the Tesla Autopilot system have brought shocking user experience and a sense of fashion and technology to reduce the burden on the driver, and partially relieved the pressure of the driver to deal with complex road conditions, and to some extent improved the driving safety. However, the automatic driving technology is still being continuously improved, and it cannot completely replace the dominant position of human beings. We cannot yet trust 100% of them. The driver must always pay attention to the road conditions.

The experience of using autonomous driving systems on different domestic roads is different. On relatively empty roads, the lane line is clearly visible and there are few frequent trips. At this time, the autopilot system is used at a low speed. The vehicle can recognize the current lane, make a smooth turn, and actively reduce the speed to keep the distance. The driving experience is quite comfortable. But in the fast-track sections of the city, especially on the high-speed roads, the feeling is different. Attention needs to be kept highly concentrated and ready to take over the steering wheel and brake pedal to deal with possible sudden problems. For example, at present, Tesla cannot recognize traffic light signals. When it meets an intersection, it will not take the initiative to decelerate the parking. At this time, artificial active braking is required. When driving at high speeds, in the event of an individual driver forcing himself in front of the line, Tesla could not recognize and respond effectively even if he had a turn signal.

In general, when driving at high speeds, automatic driving tends to make people nervous and tired. At low speeds, the pre-judgment time is longer, and it is easier to observe the surrounding traffic conditions. Through use, I personally prefer to use the low speed autopilot function.

Especially when traffic jams, the rhythm of the stoppage will make the driver feel irritable, bored and even exhausted. If the automatic driving function passes the effective fusion of various sensors and the judgment of artificial intelligence, it is possible to achieve a low-speed slow walking in a traffic jam state and liberate the driver, which is definitely a rare benefit for drivers.

Of course, China's national conditions are still quite different from Europe and the United States. Europe and the United States occupy a large proportion of roads, less road-wide vehicles, more driving across the city, the speed is naturally faster. The domestic vehicle density is relatively high, especially during peak hours of commutes. There are many people on the road in the city, and low-speed driving is the norm.

Based on the above two points, Chinese developers should focus on low-speed automatic driving, such as automatic parking, low-speed follow-up vehicles, and so on. For automakers, to properly develop the autopilot function and to make a safe, ultimate car, it is necessary to start with the assisted driving function and start with low-speed assistance.

In terms of autopilot assist functions, Xiaopeng’s solution at this stage will be a series of smart driving functions such as active cruising, lane keeping, automatic lane change, auto following or overtaking, and automatic parking. With automatic driving and holographic warning functions, we developed a low-speed automatic driving assistance technology. At the same time, due to the low speed and relatively controllable risks, we believe this is the best entry point into the field of automated driving.

Human development has always been a process of self-liberation. The history of the development of human science and technology also confirms that the development of any new technology is progressing in twists and turns. The sinking of the first giant wheel, the first space accident, and the first nuclear leak have been difficult to stop the human exploration of science and technology. We are more persistent that automatic driving is the direction in which we continue to move and will not change because of an accident.

Lei Feng network (search "Lei Feng network" public concern) Note: This article author Xia Xi, article starting cloud. Reproduced please contact the authorizing and retain the source and the author, not to delete the content.


Automatic Gas Cooker,Natural Fire Stove,Mini Cooker,Wifi Control Cooker Stove

JOYOUNG COMPANY LIMITED , https://www.globaljoyoung.com

Posted on