Tesla Inc. and its electric cars are
frequent guests at Geektimes. No wonder, because cars with autopilot function (albeit incomplete) for many of us is an amazing product of modern times. So, in fact, there are - new electric cars are more gadgets than cars. Unfortunately, for the time being everything is not so good “in the Danish kingdom”, because Ilona Mask’s cars sometimes get into
accidents .
There is no surprise - given that they travel a lot on the roads, we can only be glad that there are only a few accidents. For accidents involving ordinary cars, everyone has long been accustomed. Sometimes people are only horrified by the number of victims or the complexity of the accident, and only that. And if an electric car gets into a car accident - everyone is interested in that, yes, how. Until now, Tesla Inc has been able to prove that the perpetrators of the accident are the owners of Tesla of various modifications or other road users. But in the last accident, everything is a bit wrong.
The fact is that it says a lot about the failure in the work of Autopilot, the “digital assistant” driver, who for some reason began to steer towards the concrete separator. The electric vehicle crashed on this barrier, unfortunately, the driver also
died in the accident. The situation is so unusual that Tesla was even barred from participating in the investigation of the incident, which has never happened before.
By the way, the company takes about $ 5,000 for the opportunity to use the functions of the autopilot. Paying an additional $ 3,000, buyers get a "fully autonomous system." On this occasion, the company's website says: "All you need is to just get in the car and tell her where you are going." And then the system itself chooses the optimal path, rides through the streets, where the marking is damaged or not, identifies traffic signs and traffic light, and also does much more, which is usually the responsibility of drivers.
All this is only a description on the Tesla website. The most interesting thing is that so far there is no “complete autopilot”. The problem is that all the company's developments related to this area are at the stage of validation and consideration by the regulator. It is clear that to say exactly when the autopilot, and not just a digital assistant driver appears in Tesla cars, it is impossible.
In addition, not all functions are ready, that is, much is at the stage of development or even design. It may take years before the promised technology appears on the market. And then - not all experts on AI and neural networks have the confidence that a truly reliable autopilot (and not even a partial one) will really appear in the near future.
"This is a much more difficult problem than most people imagine," said Sam Abouelsamid, an analyst at Navigant Research.
In principle, Tesla has already taken money for products that have not yet been marketed, all we have to do is recall Tesla Model 3. Long before the start of sales, the company announced the collection of pre-orders for these cars. At first, everything was not shaky or shaky, the machines were produced, but more slowly than planned, the lag on the schedule could not be made up until now. However, buyers still started getting what they wanted. But with Autopilot (its full version) the situation is somewhat different.
Some experts generally doubt that Tesla cars without substantial modifications will be able to ride completely autonomously, without the participation of the driver. The problem is that there is no lidar in the system. It is in the Waymo robobiles (this is a subsidiary of Google), Uber, GM. But Tesla has no lidar. According to Ilona Mask, it was necessary to abandon this element for a number of reasons. So the engineers left only the radar and cameras.
Many developers of autonomous machines who consider the lidar to be one of the most important elements of a rover mobile disagree with Mask. The fact is that the same cameras can give a decent “picture”, but they cannot tell anything about the distance to the object (or rather, they can, but the accuracy will be low). Also, in case of poor lighting, the camera is not a very reliable assistant. One of the accidents associated with the cameras, has already happened - last year, Tesla's electric car did not recognize the van passing by, and crashed into it at full speed. This also led to the death of the driver. True, his guilt was proven - the fact is that the owner of the electric car did not keep his hands on the steering wheel, as required by the instructions. But if there was a lidar, maybe it would have turned out differently, and no accident happened.
The human driver also has no lidars, we use our eyes and the brain that processes visual information. As a result, a person understands much of what he sees on the road, and more or less successfully navigates, promptly responding to the appearance of external factors. There are, of course, problems, but not as often as it could be. And all thanks to millions of years of evolution, which made Homo sapiens sapiens the owner of the mental apparatus and a more or less sensitive system of smell, sight, and touch. With their help, we are guided in the world around us.
But Tesla's autopilot is not as well guided as it would be desirable for both the company's representatives and its clients.
Technology availability
The Society of Automotive Engineers (SAE) has developed a five-point autonomy scale for vehicle control systems. “0” is an ordinary car without any cruise controls or other smart stuff at all. “1” is the most basic system that may be able to help park. Well, "5" is a completely autonomous system that does not require the participation of a motorist. In the latter case, a person can listen to music, read newspapers, sleep, without paying any attention to the controls of the machine.
Autopilot Tesla is difficult to estimate from this point of view, since the company has a vehicle management system and literally every day it “finishes” it. Each major update adds some new features that enhance the old toolkit.
But here lies the danger. The fact is that in cars with a range of autonomy from “0” to “1”, the driver understands that everything depends on him and pays all his attention to the road (of course, this is in an ideal situation). At the level of development of technology "5", the driver generally can not strain, this stage provides for complete autonomy, so complete that the car may not even have a steering wheel.
But with the intermediate stages of development, everything is more complicated, especially if we consider that Tesla's digital assistant is already called Autopilot. Being deceived by the name and numerous functions (which, however, work, we must give Tesla its due), the drivers of the electric vehicles of the company Ilona Mask get used to the fact that the system is able to drive the car for some time and begin to go about their business. What happens in the end - we all know. They do not pay attention to the warning signals of their electric cars and go about their business. Someone carries more, someone less, and then the car gets into an accident.
Google went a little different way than Tesla. The corporation began to develop a fully autonomous machine at the start, for which a subsidiary,
Waymo, was created. The company understood the prospects of this direction and continues to move. True, according to the laws of the United States and Europe, the corporation still had to leave the steering wheel in the car (initially, the engineers wanted to get rid of it), because the company cannot go against the rules and regulations adopted in the USA. Now Google is planning to develop an autonomous taxi that does not need a driver in order to take the passenger from point "A" to point "B". Moreover, Google is not stuttering about the commercialization of technology.
But Tesla - not only "stutters", but also sells, and not a full-fledged product, but air. It is also worth remembering that Google is a company with large amounts of cash that can afford long-term research. In the case of Tesla, everything is not so clear. Yes, the company's shares are listed quite high, she herself has a good capitalization. But Tesla Inc still remains a startup in need of money. This forces the company to look for partners, as well as sources where you can borrow money.
If something goes wrong with the promised autopilot, the shares can fall dramatically, and Tesla Inc does not have a reserve. So a relatively small nuisance can bring down the company's stock and force it to abandon its development plans for anything.
And the plans were ambitious. For example, Musk predicted the appearance of a fully autonomous vehicle control system, and 2017 was featured in his prediction. He claimed that it was this year that a person would be able to drive an autonomous electric car from Los Angeles to New York. But, unfortunately, so far this has not happened.
Now, when the previously announced time has passed, there are no signs that the company has come close to solving this problem. But advertising "full autopilot" began in the fall of 2016.
As mentioned above, before the business model worked because Musk announced the imminent launch of some new product after its prototype appeared. In the case of the autopilot, everything is wrong - there is no prototype, so it only remains to wait and hope. And well, if there is a means to wait - many motorists simply do not stand a long queue for Tesla electric cars and give up.
So far it is impossible to say when a full-fledged autopilot will appear at least approximately, and what it will represent. But Tesla continues to sell the air, and earns it (as far as can be judged) very well. I would like to hope that the autopilot of the company can be improved, but in this case the hope is not too great. In the meantime, some buyers continue to pay $ 3,000, receiving almost nothing in return. Yes, of course, everything changes, the autopilot continues to improve. But before its appearance, there are still a few years, during which market conditions can change dramatically. And in this case, Tesla is not the best position. Perhaps something will change in the future, but obviously not in the near future.