In the previous series:. Of unsolvable problems and bad avtopilotirovaniya sensors systems prevent collisions
Over the past month, little has changed. Tesla still have accidents in avtopilotirovaniya mode, this time the first accident occurred on autopilot in China.
From the comments to the previous text Elon Musk did not recognize that it needs to finished algorithms. Therefore, I decided to add more cameras and radar to improve avtopilotirovaniya. Write him someone.
That's what happened:
It is difficult not to notice similarities with the May accident:
Guess the reason for accidents is simple. It is a lack of information, namely:
- The angle of the radar survey on the center of the car is too narrow to see cars on the edges of the strip
- Use the camera is impossible to understand the dimensions of static. barriers and dynamic vehicle corridor did not understand.
While the autopilot - a set of crutches with which solve particular traffic situation. So, they operate as a road band inseparable whole. The band is busy or free whole obstacle can not take part of the band to the left or right. Obstruction can not take part of the bands at the top (see the last accident with a trailer). One of the many assumptions that constrict the task of piloting and make it somehow solved.
We use one still floating in the minds of the idea, we will make a way, that only go autopilots. We include them ITS (intelligent transport system). Go over the possible scenario, it seems it will work.
- The car knows that broke.
- Car informs intelligent transport system that stood in the left lane. Technical assistance has already left.
- ETS puts a label of the accident in this place, and recommends moving back to readjust to the right.
- Cars took the information, rebuilt or go another route If the points?
1. The car can understand that his failure prevents movement. With the exception of accident (when the airbags are triggered) man confirms the call to the operator. So already working ERA-GLONASS and eCall.
All right, put a tick.
2. The vehicle transmits its coordinate and the fact of breakage. Report what part of the road he took, while it is impossible. Navigation is not enough, and the car - not a material point, can get in the way or at an angle. Detailed information can be understood by the cameras or radars of the road infrastructure, but in this case they have the whole road should be closed.
Tick is not set to fully cover even the urban road is still very far.
3. The signal from the vehicle is received, the damage inflicted on the place grid road. ITS has sent information to other vehicles. Information Protocol is, provide information even in twenty different formats to different vehicles is not a problem.
If you have not yet decided what will be. Checkmark.
4. Get information about that in determining the coordinates of an obstacle is not a problem, but what about the rest?
4a. Every motor vehicle shall be equipped with a navigation system. Just to compare your route and the problem on the way.
Not decided navigation feature at least 20% of all cars.
4b. The car does not know which lane is moving, navigation accuracy is insufficient. It is necessary to resort to indirect measurements (left if cars are going to meet you in the left lane), reconciliation for Beacon (can stand on the separator) or by comparing the data objects with high precision map. Most likely we can expect a combination of all three options.
It does not work.
4c. Rebuilds into free strip and get up when it is impossible to continue the movement. The autopilot must understand he will be able to pass in free space or not. Infrastructure does not have the necessary information required accuracy.
It does not work!
As a result, practical implementation of such an elementary script as a detour car broke down, resulting in the need to completely hang all the way radar or cameras, additional navigation marks, not to mention the markup and all ostalnym.I still need a car will have its own sensors to navigate.
And this is no problem. Just such a clever way to autopilot-nuzh-on!
ROAD INFRASTRUKTURAPochemu all so invested in self-managed car? Because the autopilot is not necessary to build the infrastructure. All have been built and it is a public road.
From a technology standpoint hundred times easier to automate trains, trams and underground, in which one degree of freedom within the rail. Works are conducted, but even in terms of the use of collision avoidance systems vehicles at least ten years ahead of rail.
Dedicated road road (to place emphasis on the taste), or simply impossible. Imagine a dedicated rail network to the entrance to each. Three-dimensional space is not enough.
Therefore, no one is invested in autopilot-only lines or dedicated lanes. It is useless in terms of economy projects. speed control, count cars, chat with each other? Another thing
Conclusion:. Dedicated lanes for autopilots not wait all relish the autopilot to self-sufficiency
V2XIdeya give the car more information about the road is good and technically feasible. Speaking of family V2X interfaces typically involve V2I (vehicle infrastructure) and V2V (vehicle-to-machine). There are other cases, like V2B (car-bike) or V2P (car-pedestrian).
At first glance, everything is good with V2X. With low-level protocol agreed. Several private applications, such as travel costs, too, agreed. All who use transponders on the M4 and other toll roads, using V2I.
High-level decisions every automaker says he and others are not divided. Hence incompatible standards, which hamper the development of V2X. Around the fools who do not understand the benefits of open standards? It is unlikely that this is the case. All are among the fuel nozzle and tire sizes too universal.
I suppose that car manufacturers are afraid of "commoditization" of transport, that is, the transformation in anonymous Car Trolleys, which are selected without regard to the brand. Then you compete only one price, which will burn quickly market. Consumers will enjoy a short time, the fall in profits should be a sharp drop in the budgets for the development. Machines will become worse.
How does it look in reality? The most famous competition - this Euro Truck Platooning. It involves incompatible with each other trucks that are already able to ride on public roads, the convoy, with one driver for 4-5 trucks. Those who are interested in the topic, I recommend reading a fresh report to the test in 2016. It has a lot of texture, interviews and all that we love so much. For the rest of summary: still too much problems, but they can be solved, especially if we remember that platooning going to run in 2025 .... This is the question of horizons avtopilotirovaniya.
Technically, to ensure that the information given to other machines, V2X ready now. Transmit information about traffic lights, signs, speed limit - too. So:
Conclusion: V2X is already in use and will be used in the future. As an extension of the information from the card - is, as a way of sharing information realtime-cars -. In the future
RELIABILITY DANNYHProblema does not seem ambitious, but not impossible to label it. The use of incorrect data will quickly lead to a crash.
Forgery data infrastructure is already discussed in the comments, this is possible. With its own sensors better, that's a fresh experience disruption of the autopilot Tesla (http://www.forbes.com/sites/thomasbrewster/2016/08/04/tesla-autopilot-hack-crash/#45e4b38edc93). Camera diodes burn, but she manages. Radars are jammed expensive special equipment, but only static. A simple ultrasound shit :).
So the basis of the car will take its own radars and cameras, which partly overlap, to take into account information from V2X and make a decision. At the same time giving anonymous data back into the ITS, so that she compared the information from the vehicles with their own. Good feedback channel.
What will be used to improve the accuracy and detail of the information?
- The data from the car navigation.
- Inertial sensors in the car on the road and Beacon (a fruit company has patented its recently).
- The data from cameras and radar guns Autopilot will always have their own sensors and tricky system of decision-making, taking into account the reliability of different sources.
MORAL AND LEGAL ASPEKTYDopustim, technically we have plus or minus ready to avtopilotirovaniyu. And morally? I think not: autopilots action began to discuss publicly a year or two ago
Main topics for discussion and the loss of two seats in the truckers is not among them.
First, let's understand that the autopilot will kill any people not indirectly, but directly in the accident. What is human? KAMAZ questionnaire revealed that there will be willing to sentence a person to save the dog's life. Almost 3/4 of respondents believe that the autopilot must kill his passenger, rescuing a group of people. It is true? Exactly when he is not sitting in bespilotnike respondent. At the cost of his life to save others are willing to less than a quarter.
Most pulp begins next. Wonderful MIT moral machine (go survey is short-lived) shows that we are ready to delegate the choice of victim autopilot, taking into account its value to society. Simply put, it is better to knock the old criminal, than the young prodigy. Click the link above for more detail, but given some of the racial characteristics of crime autopilots have to kill black five times more often than whites. I would not want to explain it to the stage at the presentation of the new model BMW.
Secondly, the autopilot is still reduce the total loss of humanity from the accidents. The number of victims will decline by 30-90 percent (yes, estimates vary). It is in any case a lot, but the number of victims of accidents in developed countries, and so reduced by half in a decade, what autopilots are irrelevant.
Attention to the question. How many accidents do we consider acceptable for the autopilot? Studies on this point, but most likely approach will continue. "On average," people satisfied with even 10-percent reduction in the number of victims, but only if the autopilot will not kill them precisely relative. Once the account autopilots victims will go to hundreds and thousands (a matter of time), we are waiting for the discussion non-acidic degree.
Since I addressed the issue of liability. Respond to it globally will be able to define robotic personality that Europeans even offer to discuss. Not right now, but on the horizon of 50 years. Autopilots have gone not much further. US NHTSA as it allowed an unmanned driving, but the same desire to see from the California driving drone operator has not given up.
By the way, as an idea: the issue of liability can be converted into a monetary plane. By reducing the total number of accidents affiliated with automakers insurance will pay huge compensation to the relatives of the victims, which will reduce the level of negativity
Conclusion:. Before a full understanding of the moral and ethical implications of the launch autopilots we are still very far away, well at least begun to discuss <. br>
EKONOMIKAPoslednyaya topic, it is also the most important. Still, the whole business - about money
Autopilots have a positive economy, marked on a global scale. Large companies and funds available to invest billions in R & amp; D, because it all will pay off handsomely. Morgan Stanley believes that with 100 per cent for distributing road autopilots save the world about 5, 6 billion dollars annually. This is 5% of global GDP by that time, had not toys.
Dubbed by some of the economic consequences avtopilotirovaniya
What all agree:. Fuel companies will suffer. As part of the US economy will make up to $ 160 billion. Together with the spread of elektrmobiley is an important factor of pressure on oil prices and the different energy superpower. Yes, sheikhs read this report Morgans.
Transportation will be cheaper and easier. The cost of transportation of three quarters - is the work of the driver. Simplify intermodal transport, hence the attention to start-ups in this area. Globally, the cost of transport will decrease approximately by half, and increase their profitability. The cost of the release of 90 percent of drivers, sorry. Is there any yogurt will fall in price by a percentage. And the price of sand or concrete delivery of more than 50%, that is, so the whole building will fall in price in the world.
There will be fewer accidents and hospital bills. Savings estimated at 180-190 billion dollars in the US alone. By the way, this amount will fall in GDP and the same money will not receive service companies and medical institutions
Implicit points:. Automakers - the current business model rather lose, cars need less time. But they will be given more services. Hence the attention to any auto-start-ups, auto giants pillars place. To everyone understood car market - about $ 2 trillion a year, which have and continue to grow
In humans it is released about 50 minutes of time in the day. Most likely this will benefit manufacturers and infotainment alcohol. The press, of course, all write about improving the performance and once an hour a day, which can be spent on the work.
The release areas from roads and parking lots - and it is a land mainly in the cities. 5, 7 billion square meters is difficult to calculate in money. Vehicles should be cheaper.
We solve the important problem of the last mile in the delivery of goods. It is important and will affect the insane number of companies. For example:
"business model: 25-30% commission on food and 2.5 euros for a delivery to the customer. It sounds profitable. But actually no, because couriers must be paid at the rate of 15 euros / hour, which at the current average of the check should be more than 1.5 deliveries per hour.
The previous two rounds of money with investment total of 16 million euros over, and the need to "density" of deliveries did not score. The third round of investment promised, but then withdrew its proposal dyudila back.
1. Logistics is able to gobble up all the money. "We can wait for the diffusion of new services, which do not even think right now. The most simple - it is a cheap (around $ 1) delivery of croissants in the morning, thousands of homes
Instead output: Consensus on the future of automakers and insurance companies do not have.. It seems to be quite good report Munich RE (one of the largest reinsurance companies, 130 years, 280 billion euros of assets) in which much has been written and said nothing. So in advance Respect experts who kamentah raspishut in three lines, what will happen to the market.
- People are just starting to think about how the automatic pilots will evolve in existing financial, yurdicheskie and many other aspects of life. No particular hurry, 15-20 years, we have
- knowingly Avtopilotiruemy car goes so complicated -. It is programmed "Gordian knot", which merged all the problems related markets. For example, infrastructure.
- The effects of the use of autopilots is difficult to understand until the end, so strong and different influences avtopilotirovanie on dozens of spheres. If you are interested in my opinion - as well as other robotics, it brings humanity to the unconditional income.
- Oh, and forget about has finally selected the road.