[Python, IT-инфраструктура, Big Data] You are standing at a red light at an empty intersection. How to make traffic lights smarter?
Автор
Сообщение
news_bot ®
Стаж: 6 лет 9 месяцев
Сообщений: 27286
Types of smart traffic lights: adaptive and neural networksAdaptive works at relatively simple intersections, where the rules and possibilities for switching phases are quite obvious. Adaptive management is only applicable where there is no constant loading in all directions, otherwise it simply has nothing to adapt to – there are no free time windows. The first adaptive control intersections appeared in the United States in the early 70s of the last century. Unfortunately, they have reached Russia only now, their number according to some estimates does not exceed 3,000 in the country.Neural networks – a higher level of traffic regulation. They take into account a lot of factors at once, which are not even always obvious. Their result is based on self-learning: the computer receives live data on the bandwidth and selects the maximum value by all possible algorithms, so that in total, as many vehicles as possible pass from all sides in a comfortable mode per unit of time. How this is done, usually programmers answer – we do not know, the neural network is a black box, but we will reveal the basic principles to you…Adaptive traffic lights use, at least, leading companies in Russia, rather outdated technology for counting vehicles at intersections: physical sensors or video background detector. A capacitive sensor or an induction loop only sees the vehicle at the installation site-for a few meters, unless of course you spend millions on laying them along the entire length of the roadway. The video background detector shows only the filling of the roadway with vehicles relative to this roadway. The camera should clearly see this area, which is quite difficult at a long distance due to the perspective and is highly susceptible to atmospheric interference: even a light snowstorm will be diagnosed as the presence of traffic – the background video detector does not distinguish the type of detection.
Neural network traffic lights also use neural networks for object recognition, which gives them a more informative and super accurate picture of what is happening. These "brains" recognize any images in their any positions and therefore more accurately count the number of vehicles, do it at a much greater distance from a single camera. In addition, neural networks determine the qualitative composition of the vehicle both by type and by a variety of other parameters. For example, a long truck and a fast Ferrari have a completely different way to start driving, which affects the algorithms of smart traffic lights in different ways. Even the blonde with the "beginner" sign on the vehicle, who constantly forgets to start at the green light is gradually included in the coefficient of correction of calculations. In addition, neural networks are able to recognize people, which is also very important – because pedestrians also want to pass through the road. Neural networks also diagnose more than a hundred other objects, from baby strollers with bicycles to suitcases and weapons.
We also note that adaptive traffic lights are very demanding to the rules for installing video cameras, the height of the suspension, the lack of overlap in the view, which is quite difficult – there are always some wires or poles blocking the view. As a result, installation becomes more expensive, it is rarely possible to use existing supports – you need to build special farms. Neural networks generally have almost no restrictions on the installation points, they recognize any objects from any angles in any form and with large overlaps-often it is enough to see only 15% of the car to understand that it is a car. Accordingly, the neural network traffic light is cheaper in terms of installation, in the number of video cameras, the length of wires and other installation tricks. We cannot say about the complexity of adjusting cameras for adaptive traffic lights, their video detector is rigidly tied to the type of roadway. As a result of winds and vibrations from trucks, these video cameras constantly lose their view and require almost weekly adjustment, and this is the "tower", people, and costs. And traffic restrictions for the duration of work. Neural network cameras will understand everything that is happening as long as they somehow see the road situation, even large shifts are not critical for them. Let's take a closer look at where and how these smart traffic lights can increase throughput: Example 1. a secondary road Adjacent to the highway.
When only one car passes in a hour from a nearby village, we are obliged to let it pass anyway. But to do this, we will have to stop traffic every 5 minutes – it can't wait an hour. If it is enough for a pedestrian to hang a button to turn on the green, then the car does not have this option. In this configuration, both types of traffic lights are equally suitable: adaptive – because this situation is absolutely obvious, neural network-because it is cheaper. (Why it is cheaper-we will analyze further.) Someone here will say: - Let's build a speeding lane-and the traffic light is not needed at all. - but cars sometimes need to turn left – then you will have to make ring for turning cars. And we also have pedestrians who are also not averse to crossing the highway – they need to build an underground or aboveground crossing. All this will require tens of millions for construction and millions for maintenance. To put and maintain two poles with three light bulbs and a pair of video cameras-the figures are two orders of magnitude cheaper. Although ... there is a choice. Example 2. a city intersection at an unloaded time.
In ordinary cities, we often encounter a situation where we stand at a red light while there is no one in the transverse direction. For example, in Ivanovo, the rush hour is only a few hours a day. And at night, and in general, all of us are unnerved by the unnecessary waiting at almost every intersection. Why do we waste time and gas and wear out the brake pads? If there is no traffic in any of the directions or it quickly ended after turning on the green signal there, the traffic light has all the prerequisites – to give passage to the loaded section. But we are stoically waiting for our phase! No schedules can solve this issue, the usual logic that is in any traffic light object controller, will be wrong specifically on this day and at this time. Although manual adjustments, of course, are carried out and they bring some effect. But it is much more correct to have feedback here and now: if the video cameras see where the transport is going, then it is necessary to turn on the green in this place. And, in advance-without forcing to slow down. Again, both types of traffic lights are suitable: adaptive-because the situation is obvious, neural network is slightly better-because it sees vehicles further, you do not need to brake, because we need to give a command for a long time – it will take a few seconds for yellow. In addition, we have again there are people. Adaptive will constantly give them some time to transition, because it does not distinguish between them. Neural network – will switch only when it appears. You again wanted to say something: pedestrians can be given a button. Yes, but in some way, it's not kosher in the presence of a smart device. In addition, you need to determine whether everyone has moved, and this can only be done by a neural network traffic light. Adaptive will only give you the average time to move at a fast pace. and the situation here is twofold: some little boy quickly ran across, and some grandmother will hobble for a long time. Or, for example, will go young army pupils from nearby school. You also wanted to say something: at night, you should turn on the flashing yellow light at all intersections! - we have tried it for many decades. We found out, that in our country it is very dangerous. The Internet is just full of car racing videos, no speed meters stop night lovers to accelerate on city roads up to 300 km / h. It's not just cars that turn into trash, people also cross roads at night. And, if we are talking about night mode, neural network traffic lights are much more efficient not only in terms of bandwidth – you do not have to brake if you are driving at normal speed, they also perform the role of a police officer – they can stop any rider. Neural networks determine the speed of rushing, the presence of other vehicles and pedestrians who may be threatened by it, and calculate the degree of danger. And, specifically, we are against turning a smart device into a punitive mechanism. if the computer sees even increased speed, but does not find a threat, then specifically our logic does not prevent the passage of, let's say, the violator. This is not done, because we somehow do not recognize the law, we have set up our equipment as a reserve for the future. From our point of view, strict speed limits, in principle, are not necessary. To whom with what speed and in what place to move-the computer should determine! He calculates everything and gives a fairly accurate solution. Specifically, our smart traffic light stops only those whose speed really threatens other road users. And catching violators is another topic, there are special devices for this.Example 3. Constantly busy intersection.
Here, probably, everyone understands that only a neural network traffic light is suitable, because there is simply no obvious logic for switching traffic light phases: there is a continuous flow of vehicles from all sides, and who should be given priority? However, we need to allocate some time phases, they have a length. How much time should I give for traveling from North to South, so that the maximum number of vehicles passes? One minute, two, three? Or maybe 569 seconds? Who should say this, some very smart person? Or apply a scientific approach? We haven't told you anything yet, but we've already heard your question: "What does it matter? – No, because the movement is inhomogeneous, we have a motor vehicle spring: when the flow starts, it stretches, when it stops, it contracts. And by time often exceeds the process of movement between neighboring intersections. The more stops, the less the vehicle will pass as a result of all the time. And you, immediately suggest: - Then you need to give the longest phase possible, so as not to create frequent springs. – It would be correct ,if we had no other direction: from East to West. It also requires a maximum phase. This game is called "wolves-sheep": Those, who go by car-are wolves, and those who stand -are sheep. And you, without thinking, will say: - Let's give everyone equally! - Let's say! But how much is equal? Okay, we agreed for 5 minutes each. But one direction is stuck in the second intersection, and these 5 minute -pointlessly, it is traffic jam after 3 minutes. The other two minutes, the intersection is just clogged with barely moving cars. We lose traffic speed, which also leads to a decrease in bandwidth.As a result, we must also take into account the speed of all streams at any given time. By the way, the compression time of the springs in each direction also depends on it: the higher the speed was, the longer the flow stop. Also, the springs depend on the type of vehicle: the truck will take much longer to accelerate, the passenger cars are much faster, so we need to understand the type of vehicle and its dimensions.Извините, данный ресурс не поддреживается. :( Loading by loading, but often there is another pronounced picture: in the morning everyone goes in one direction, and in the evening in the opposite direction. At the same time, very often there is an almost free band. And someone can be skip on it from those that go in a perpendicular direction, thereby freeing the "sheep" from captivity in advance, while most of them are standing. Turning on the arrows - there can be a lot of them. But then pedestrians are outraged – they also want to go. And in the other direction, there are no pedestrians at all. And one person fell while crossing, and severely injured his leg, remaining on the roadway. He needs help, and the usual traffic light out of habit lets vehicles go there: accident, traffic jams again! By the way, about accidents, this is still a whole complex of prefatory with a thousand unknowns. Calculating the optimal modes of all traffic light objects is a task for a computer, not a primitive everyday logic. When a truck dropped a crate at a perfectly simple unloaded intersection, the adaptive traffic light went crazy. He constantly trying to skip the loaded direction, and that was physically impossible. As a result, another – free direction-got into the traffic. And the neural network traffic light calculates millions of constantly changing parameters, selecting the optimal algorithms of operation, as a result, it provides an increase in throughput in any, including emergency conditions, and also increases safety: it automatically stops transport in lanes ,where one collision can develop into a serial one. And this is only a 30% increase in efficiency at always busy intersections. The most interesting thing is that this figure is easy to bring to 60%, how do you like it? Not tired of standing in traffic jams? Want to drive twice as fast? It has already been said that the most expensive part in this process is the motor vehicle spring: it takes somewhere UP to, and somewhere MORE than 50% of the time to stop and start. But if we don't stop more than half of the vehicles? If connect neighboring Smart traffic lights to our Smart traffic light, the neural network gets valuable information: when, where and at what speed the flow is moving, what type of transport in each flow, even the driving style of individual vehicles (remember the blonde that everyone goes around?). Thus, we can already remove every second "spring", or even more. - We see that you also have constructive suggestions: - And what if the pre-previous traffic lights will be also included in the General system? – That's right, this way you can drive around Moscow or St. Petersburg many times faster!What about now? In large cities, smart automatic traffic control centers have been built, in which professional specialists (sorry for the tautology) manage traffic light objects. They, of course, do not press the button for each light bulb, but it is manually set the operating modes for each traffic light. In other cities not even that -just the average schedule. Perhaps, it's cool for 1930 years! Not today, but the last century, when a traffic controller stood in a booth on the street – now his work is carried out remotely. Although..., a person can beat a computer at chess! There are also a million unknowns in this game. Perhaps in the automatic traffic control system are working pros. But how many grandmasters are there in a million people who just know how to play this game? Still, in 99.999999999999 ... % will win the computer. After these words, there is usually a heavy pause – people do not like to believe in some kind of artificial intelligence, it is too often called completely unintelligent things in our country. Therefore, we will give an example of a completely obvious efficiency, which is understandable to the living intellect. Lets take three consecutive intersections in both directions, located at a short distance from each other – say, a minute's drive away. If we turn on the "green wave" in each direction, then both will miss transport in 3 minutes. Note, that this is a comfortable waiting time for "sheep". But now we have a spring: two starts and three brakes, which triple the time of each flow and, accordingly, the waiting time for the transverse one. 3 times higher throughput at busy intersections!And again the question: - May be you will make 3 sections where transport will fly, but all this flow will unite with the fourth – not so smart-intersection! - Let me immediately disagree with the phrase " all this flow". Congested intersections are located in cities, usually in their centers, respectively, there are a lot of residential buildings, offices, shops and other places where people go. Accordingly, for 3 of sections the flow will not stop at the 4th intersection, some will resolve, the neural network control system can easily calculate this. In addition, during this green phase, we exclude the influx of other cars, thus we can determine the maximum amount the vehicle and, accordingly, bring all to this stretch of road, which will allow you to create a reasonable queue in front of the usual intersection. Respectively, in advance need to choose the long stretches of road before the old version of traffic lights. There will be no collapse here! Although it is correctly noted, that it is better to tie the entire city to the general neural network management system. And this is also possible and necessary, we will all come to this one day! Smart traffic lights based on neural networks are the absolute future for Russia. It is impossible to endlessly build high-rises in the center of Moscow, torpedo the sale of cars and at the same time drive slower than walking. For the entire developed world, neural network technologies have already become commonplace, and such tasks are solved very quickly there. A little more about the violators. Just not speed – for us, high-speed travel is a blessing, but traffic lights. According to our mathematical model, it is remarkably clear that traffic violations reduce the throughput of any intersections by up to 6% during normal times and up to 25% during busy periods. It is clear, that everyone want to drive faster, and they go to the territory of the intersection on a forbidding traffic light signal or just behind the stop line, creating interference, or dangerously maneuver the row, so that even thе smart intersection may lose his mind. Of course, neural networks will react, and in the direction of security, but this process undermines the computational plans of computer logic, thereby bandwidth. We have studied many mathematical models and all of them suffer from traffic violations. So far, we can offer only one option to deal with violators – record video and immediately dump it on tablets in the cars of the traffic police patrol services for a preventive conversation. In the future, it is possible to compile black lists of state registration mark and a special reaction to their appearance.Integration
This is the simplest question from the point of view of technical implementation and the most difficult one from the point of view of combating technocracy. It is enough to put 4 cameras in all directions, even on existing supports, and in most cases you can start working. The main element is the software "SpeсLab-traffic light". Of course, the more cameras, the higher the effectiveness, the number will be select by designing. The video decision server can be located anywhere in your city or, if you do not want to serve it, in our office. Its presence on the traffic light object itself is not required, because it does not involve super urgent control commands, a one second- is a sufficient speed for passing commands. Even if a speeding ambulance is detected, we can't switch directions instantly – there should always be a yellow warning period. And in general, the yellow signal is very widely used by our smart traffic light for most unexpected cases. Today, drivers of emergency services are afraid to pass a red light, despite the legality of such a maneuver. However, in the event of an accident, they will be to blame. And yellow, although not permissive, significantly increases the rights of fire or ambulance. The video decision server can use an existing traffic light controller, if it is not quite ancient. Our software only gives commands to change the operating modes of this controller. In case of loss of communication, this controller will go into default mode – that's all.About the most difficult!
Today, traffic lights of the last century are already everywhere. Under them, a huge infrastructure has been created with a lot of money for maintenance. The leadership of the regions of the Russian Federation is technocratic in itself, and there is also a redistribution of the market. This task cannot be solved without political will.There is also positive news, the Russian Government is actively developing an Intelligent Transport System, at least in words. There are quite a lot of resolutions of the Government of the Russian Federation for this, moreover, money is allocated – at least according to this decree. But in our experience, progress is so strongly resisted on the ground, that they often do not even master these tools, despite the fact that there is a penalty for this. There is one more news of average positivity: Progress will still come! It would be wrong to think that the money is allocated just like that, government circles are actively moving Moscow firms that are engaged in the supply of adaptive traffic lights. And it is difficult to resist against Moscow, it will come and demolish all the" politicians " of the region. But the news is average, because adaptive management is the last century, it has been used in the United States for 50 years and today it is changing to a neural network. Thus, it is safe to say that Smart traffic lights will appear everywhere in the coming years.Start with one or the whole city at once?
We see different opinions: somewhere they want to run- in one intersection, and somewhere the leadership of the region is firmly against point changes, arguing that they will break the existing infrastructure. Believe me, arguing with the management in all cases does not make sense, but from the point of view of logic, both options are suitable. Most often, in order to do nothing, they say: there is no money for the whole city, and the gradual introduction will break the logic of other traffic lights built over the years and lead to a collapse, such as if traffic flows go faster somewhere, in other places everything will stand up. We hear it all the time. But, in cities, every day there are some changes with the capacity: roads are blocked for repairs, new streets are built, streets are narrowed in winter, new traffic lights are installed, signs are changed… Nothing has yet led to a collapse, but on the contrary-it unloads traffic. Installing a single smart intersection is equivalent to building one more lane in all directions or putting a "stop prohibited" sign on the roadsides. Such a situation is easy to calculate, and rebuild the nearest traffic lights, as it has always been done for many decades. Talk of the end of the world is greatly exaggerated. Moreover, the neural network intersection sees the traffic flow and to the exit, so it can quickly rebuild, if it understands, that there is no point in allowing to pass transport in this particular direction.For the sake of information, it should be added that in a dozen years another type of Smart traffic lights will come - Communication. Even today, Volvo cars are equipped with equipment, that can transmit data about itself to all interested traffic control systems. You don't even need cameras. But for this it is necessary, that all the old models are rusted. Although it is possible, that such devices will become mandatory - as well as airplane detector, which were introduced after one of the air crashes. Of course, useful and life-saving technologies may not reach Russia for a long time.
===========
Источник:
habr.com
===========
Похожие новости:
- [Python, Алгоритмы, Big Data, Машинное обучение, Искусственный интеллект] Data Science Digest — 21.04.21
- [IT-инфраструктура, Управление проектами, Инженерные системы] Мониторинг в ЦОДе: как мы меняли старую BMS на новую. Часть 4
- [Python, Обработка изображений, Машинное обучение, Лайфхаки для гиков] Как удалить татуировку с помощью глубокого обучения (перевод)
- [Python, Разработка под Linux, Искусственный интеллект, Natural Language Processing, Визуальное программирование] Разработчик натренировал модели OpenAI на интерпретацию человеческого языка в команды терминала
- [Информационная безопасность, IT-инфраструктура, Хранение данных, Хранилища данных] Блокировки хорошие и не очень
- [IT-инфраструктура, Apache, Big Data, DevOps] Итоговый проект для видеокурса и подкаст «Проблемная Kafka»
- [Python, Машинное обучение, Искусственный интеллект, Социальные сети и сообщества] Neural network Telegram bot with StyleGAN and GPT-2
- [Работа с видео, Python, Обработка изображений, Машинное обучение] Распознавание маски на лице с помощью YOLOv3 (перевод)
- [Big Data, Интернет-маркетинг] Машинное обучение, big data и персонализация сайта: как eCommerce поднимает продажи при помощи технологий
- [IT-инфраструктура, Виртуализация, Big Data] Вы нервно стоите на красный, хотя машин нет. Как сделать светофор умнее, а проезжать и загруженные перекрестки быстрее?
Теги для поиска: #_python, #_itinfrastruktura (IT-инфраструктура), #_big_data, #_smart_traffic_lights, #_smart_city, #_traffic_light, #_python, #_itinfrastruktura (
IT-инфраструктура
), #_big_data
Вы не можете начинать темы
Вы не можете отвечать на сообщения
Вы не можете редактировать свои сообщения
Вы не можете удалять свои сообщения
Вы не можете голосовать в опросах
Вы не можете прикреплять файлы к сообщениям
Вы не можете скачивать файлы
Текущее время: 22-Ноя 11:01
Часовой пояс: UTC + 5
| Автор | Сообщение |
|---|---|
|
news_bot ®
Стаж: 6 лет 9 месяцев |
|
|
Types of smart traffic lights: adaptive and neural networksAdaptive works at relatively simple intersections, where the rules and possibilities for switching phases are quite obvious. Adaptive management is only applicable where there is no constant loading in all directions, otherwise it simply has nothing to adapt to – there are no free time windows. The first adaptive control intersections appeared in the United States in the early 70s of the last century. Unfortunately, they have reached Russia only now, their number according to some estimates does not exceed 3,000 in the country.Neural networks – a higher level of traffic regulation. They take into account a lot of factors at once, which are not even always obvious. Their result is based on self-learning: the computer receives live data on the bandwidth and selects the maximum value by all possible algorithms, so that in total, as many vehicles as possible pass from all sides in a comfortable mode per unit of time. How this is done, usually programmers answer – we do not know, the neural network is a black box, but we will reveal the basic principles to you…Adaptive traffic lights use, at least, leading companies in Russia, rather outdated technology for counting vehicles at intersections: physical sensors or video background detector. A capacitive sensor or an induction loop only sees the vehicle at the installation site-for a few meters, unless of course you spend millions on laying them along the entire length of the roadway. The video background detector shows only the filling of the roadway with vehicles relative to this roadway. The camera should clearly see this area, which is quite difficult at a long distance due to the perspective and is highly susceptible to atmospheric interference: even a light snowstorm will be diagnosed as the presence of traffic – the background video detector does not distinguish the type of detection.  Neural network traffic lights also use neural networks for object recognition, which gives them a more informative and super accurate picture of what is happening. These "brains" recognize any images in their any positions and therefore more accurately count the number of vehicles, do it at a much greater distance from a single camera. In addition, neural networks determine the qualitative composition of the vehicle both by type and by a variety of other parameters. For example, a long truck and a fast Ferrari have a completely different way to start driving, which affects the algorithms of smart traffic lights in different ways. Even the blonde with the "beginner" sign on the vehicle, who constantly forgets to start at the green light is gradually included in the coefficient of correction of calculations. In addition, neural networks are able to recognize people, which is also very important – because pedestrians also want to pass through the road. Neural networks also diagnose more than a hundred other objects, from baby strollers with bicycles to suitcases and weapons.  We also note that adaptive traffic lights are very demanding to the rules for installing video cameras, the height of the suspension, the lack of overlap in the view, which is quite difficult – there are always some wires or poles blocking the view. As a result, installation becomes more expensive, it is rarely possible to use existing supports – you need to build special farms. Neural networks generally have almost no restrictions on the installation points, they recognize any objects from any angles in any form and with large overlaps-often it is enough to see only 15% of the car to understand that it is a car. Accordingly, the neural network traffic light is cheaper in terms of installation, in the number of video cameras, the length of wires and other installation tricks. We cannot say about the complexity of adjusting cameras for adaptive traffic lights, their video detector is rigidly tied to the type of roadway. As a result of winds and vibrations from trucks, these video cameras constantly lose their view and require almost weekly adjustment, and this is the "tower", people, and costs. And traffic restrictions for the duration of work. Neural network cameras will understand everything that is happening as long as they somehow see the road situation, even large shifts are not critical for them. Let's take a closer look at where and how these smart traffic lights can increase throughput: Example 1. a secondary road Adjacent to the highway.  When only one car passes in a hour from a nearby village, we are obliged to let it pass anyway. But to do this, we will have to stop traffic every 5 minutes – it can't wait an hour. If it is enough for a pedestrian to hang a button to turn on the green, then the car does not have this option. In this configuration, both types of traffic lights are equally suitable: adaptive – because this situation is absolutely obvious, neural network-because it is cheaper. (Why it is cheaper-we will analyze further.) Someone here will say: - Let's build a speeding lane-and the traffic light is not needed at all. - but cars sometimes need to turn left – then you will have to make ring for turning cars. And we also have pedestrians who are also not averse to crossing the highway – they need to build an underground or aboveground crossing. All this will require tens of millions for construction and millions for maintenance. To put and maintain two poles with three light bulbs and a pair of video cameras-the figures are two orders of magnitude cheaper. Although ... there is a choice. Example 2. a city intersection at an unloaded time.  In ordinary cities, we often encounter a situation where we stand at a red light while there is no one in the transverse direction. For example, in Ivanovo, the rush hour is only a few hours a day. And at night, and in general, all of us are unnerved by the unnecessary waiting at almost every intersection. Why do we waste time and gas and wear out the brake pads? If there is no traffic in any of the directions or it quickly ended after turning on the green signal there, the traffic light has all the prerequisites – to give passage to the loaded section. But we are stoically waiting for our phase! No schedules can solve this issue, the usual logic that is in any traffic light object controller, will be wrong specifically on this day and at this time. Although manual adjustments, of course, are carried out and they bring some effect. But it is much more correct to have feedback here and now: if the video cameras see where the transport is going, then it is necessary to turn on the green in this place. And, in advance-without forcing to slow down. Again, both types of traffic lights are suitable: adaptive-because the situation is obvious, neural network is slightly better-because it sees vehicles further, you do not need to brake, because we need to give a command for a long time – it will take a few seconds for yellow. In addition, we have again there are people. Adaptive will constantly give them some time to transition, because it does not distinguish between them. Neural network – will switch only when it appears. You again wanted to say something: pedestrians can be given a button. Yes, but in some way, it's not kosher in the presence of a smart device. In addition, you need to determine whether everyone has moved, and this can only be done by a neural network traffic light. Adaptive will only give you the average time to move at a fast pace. and the situation here is twofold: some little boy quickly ran across, and some grandmother will hobble for a long time. Or, for example, will go young army pupils from nearby school. You also wanted to say something: at night, you should turn on the flashing yellow light at all intersections! - we have tried it for many decades. We found out, that in our country it is very dangerous. The Internet is just full of car racing videos, no speed meters stop night lovers to accelerate on city roads up to 300 km / h. It's not just cars that turn into trash, people also cross roads at night. And, if we are talking about night mode, neural network traffic lights are much more efficient not only in terms of bandwidth – you do not have to brake if you are driving at normal speed, they also perform the role of a police officer – they can stop any rider. Neural networks determine the speed of rushing, the presence of other vehicles and pedestrians who may be threatened by it, and calculate the degree of danger. And, specifically, we are against turning a smart device into a punitive mechanism. if the computer sees even increased speed, but does not find a threat, then specifically our logic does not prevent the passage of, let's say, the violator. This is not done, because we somehow do not recognize the law, we have set up our equipment as a reserve for the future. From our point of view, strict speed limits, in principle, are not necessary. To whom with what speed and in what place to move-the computer should determine! He calculates everything and gives a fairly accurate solution. Specifically, our smart traffic light stops only those whose speed really threatens other road users. And catching violators is another topic, there are special devices for this.Example 3. Constantly busy intersection.  Here, probably, everyone understands that only a neural network traffic light is suitable, because there is simply no obvious logic for switching traffic light phases: there is a continuous flow of vehicles from all sides, and who should be given priority? However, we need to allocate some time phases, they have a length. How much time should I give for traveling from North to South, so that the maximum number of vehicles passes? One minute, two, three? Or maybe 569 seconds? Who should say this, some very smart person? Or apply a scientific approach? We haven't told you anything yet, but we've already heard your question: "What does it matter? – No, because the movement is inhomogeneous, we have a motor vehicle spring: when the flow starts, it stretches, when it stops, it contracts. And by time often exceeds the process of movement between neighboring intersections. The more stops, the less the vehicle will pass as a result of all the time. And you, immediately suggest: - Then you need to give the longest phase possible, so as not to create frequent springs. – It would be correct ,if we had no other direction: from East to West. It also requires a maximum phase. This game is called "wolves-sheep": Those, who go by car-are wolves, and those who stand -are sheep. And you, without thinking, will say: - Let's give everyone equally! - Let's say! But how much is equal? Okay, we agreed for 5 minutes each. But one direction is stuck in the second intersection, and these 5 minute -pointlessly, it is traffic jam after 3 minutes. The other two minutes, the intersection is just clogged with barely moving cars. We lose traffic speed, which also leads to a decrease in bandwidth.As a result, we must also take into account the speed of all streams at any given time. By the way, the compression time of the springs in each direction also depends on it: the higher the speed was, the longer the flow stop. Also, the springs depend on the type of vehicle: the truck will take much longer to accelerate, the passenger cars are much faster, so we need to understand the type of vehicle and its dimensions.Извините, данный ресурс не поддреживается. :( Loading by loading, but often there is another pronounced picture: in the morning everyone goes in one direction, and in the evening in the opposite direction. At the same time, very often there is an almost free band. And someone can be skip on it from those that go in a perpendicular direction, thereby freeing the "sheep" from captivity in advance, while most of them are standing. Turning on the arrows - there can be a lot of them. But then pedestrians are outraged – they also want to go. And in the other direction, there are no pedestrians at all. And one person fell while crossing, and severely injured his leg, remaining on the roadway. He needs help, and the usual traffic light out of habit lets vehicles go there: accident, traffic jams again! By the way, about accidents, this is still a whole complex of prefatory with a thousand unknowns. Calculating the optimal modes of all traffic light objects is a task for a computer, not a primitive everyday logic. When a truck dropped a crate at a perfectly simple unloaded intersection, the adaptive traffic light went crazy. He constantly trying to skip the loaded direction, and that was physically impossible. As a result, another – free direction-got into the traffic. And the neural network traffic light calculates millions of constantly changing parameters, selecting the optimal algorithms of operation, as a result, it provides an increase in throughput in any, including emergency conditions, and also increases safety: it automatically stops transport in lanes ,where one collision can develop into a serial one. And this is only a 30% increase in efficiency at always busy intersections. The most interesting thing is that this figure is easy to bring to 60%, how do you like it? Not tired of standing in traffic jams? Want to drive twice as fast? It has already been said that the most expensive part in this process is the motor vehicle spring: it takes somewhere UP to, and somewhere MORE than 50% of the time to stop and start. But if we don't stop more than half of the vehicles? If connect neighboring Smart traffic lights to our Smart traffic light, the neural network gets valuable information: when, where and at what speed the flow is moving, what type of transport in each flow, even the driving style of individual vehicles (remember the blonde that everyone goes around?). Thus, we can already remove every second "spring", or even more. - We see that you also have constructive suggestions: - And what if the pre-previous traffic lights will be also included in the General system? – That's right, this way you can drive around Moscow or St. Petersburg many times faster!What about now? In large cities, smart automatic traffic control centers have been built, in which professional specialists (sorry for the tautology) manage traffic light objects. They, of course, do not press the button for each light bulb, but it is manually set the operating modes for each traffic light. In other cities not even that -just the average schedule. Perhaps, it's cool for 1930 years! Not today, but the last century, when a traffic controller stood in a booth on the street – now his work is carried out remotely. Although..., a person can beat a computer at chess! There are also a million unknowns in this game. Perhaps in the automatic traffic control system are working pros. But how many grandmasters are there in a million people who just know how to play this game? Still, in 99.999999999999 ... % will win the computer. After these words, there is usually a heavy pause – people do not like to believe in some kind of artificial intelligence, it is too often called completely unintelligent things in our country. Therefore, we will give an example of a completely obvious efficiency, which is understandable to the living intellect. Lets take three consecutive intersections in both directions, located at a short distance from each other – say, a minute's drive away. If we turn on the "green wave" in each direction, then both will miss transport in 3 minutes. Note, that this is a comfortable waiting time for "sheep". But now we have a spring: two starts and three brakes, which triple the time of each flow and, accordingly, the waiting time for the transverse one. 3 times higher throughput at busy intersections!And again the question: - May be you will make 3 sections where transport will fly, but all this flow will unite with the fourth – not so smart-intersection! - Let me immediately disagree with the phrase " all this flow". Congested intersections are located in cities, usually in their centers, respectively, there are a lot of residential buildings, offices, shops and other places where people go. Accordingly, for 3 of sections the flow will not stop at the 4th intersection, some will resolve, the neural network control system can easily calculate this. In addition, during this green phase, we exclude the influx of other cars, thus we can determine the maximum amount the vehicle and, accordingly, bring all to this stretch of road, which will allow you to create a reasonable queue in front of the usual intersection. Respectively, in advance need to choose the long stretches of road before the old version of traffic lights. There will be no collapse here! Although it is correctly noted, that it is better to tie the entire city to the general neural network management system. And this is also possible and necessary, we will all come to this one day! Smart traffic lights based on neural networks are the absolute future for Russia. It is impossible to endlessly build high-rises in the center of Moscow, torpedo the sale of cars and at the same time drive slower than walking. For the entire developed world, neural network technologies have already become commonplace, and such tasks are solved very quickly there. A little more about the violators. Just not speed – for us, high-speed travel is a blessing, but traffic lights. According to our mathematical model, it is remarkably clear that traffic violations reduce the throughput of any intersections by up to 6% during normal times and up to 25% during busy periods. It is clear, that everyone want to drive faster, and they go to the territory of the intersection on a forbidding traffic light signal or just behind the stop line, creating interference, or dangerously maneuver the row, so that even thе smart intersection may lose his mind. Of course, neural networks will react, and in the direction of security, but this process undermines the computational plans of computer logic, thereby bandwidth. We have studied many mathematical models and all of them suffer from traffic violations. So far, we can offer only one option to deal with violators – record video and immediately dump it on tablets in the cars of the traffic police patrol services for a preventive conversation. In the future, it is possible to compile black lists of state registration mark and a special reaction to their appearance.Integration This is the simplest question from the point of view of technical implementation and the most difficult one from the point of view of combating technocracy. It is enough to put 4 cameras in all directions, even on existing supports, and in most cases you can start working. The main element is the software "SpeсLab-traffic light". Of course, the more cameras, the higher the effectiveness, the number will be select by designing. The video decision server can be located anywhere in your city or, if you do not want to serve it, in our office. Its presence on the traffic light object itself is not required, because it does not involve super urgent control commands, a one second- is a sufficient speed for passing commands. Even if a speeding ambulance is detected, we can't switch directions instantly – there should always be a yellow warning period. And in general, the yellow signal is very widely used by our smart traffic light for most unexpected cases. Today, drivers of emergency services are afraid to pass a red light, despite the legality of such a maneuver. However, in the event of an accident, they will be to blame. And yellow, although not permissive, significantly increases the rights of fire or ambulance. The video decision server can use an existing traffic light controller, if it is not quite ancient. Our software only gives commands to change the operating modes of this controller. In case of loss of communication, this controller will go into default mode – that's all.About the most difficult! Today, traffic lights of the last century are already everywhere. Under them, a huge infrastructure has been created with a lot of money for maintenance. The leadership of the regions of the Russian Federation is technocratic in itself, and there is also a redistribution of the market. This task cannot be solved without political will.There is also positive news, the Russian Government is actively developing an Intelligent Transport System, at least in words. There are quite a lot of resolutions of the Government of the Russian Federation for this, moreover, money is allocated – at least according to this decree. But in our experience, progress is so strongly resisted on the ground, that they often do not even master these tools, despite the fact that there is a penalty for this. There is one more news of average positivity: Progress will still come! It would be wrong to think that the money is allocated just like that, government circles are actively moving Moscow firms that are engaged in the supply of adaptive traffic lights. And it is difficult to resist against Moscow, it will come and demolish all the" politicians " of the region. But the news is average, because adaptive management is the last century, it has been used in the United States for 50 years and today it is changing to a neural network. Thus, it is safe to say that Smart traffic lights will appear everywhere in the coming years.Start with one or the whole city at once? We see different opinions: somewhere they want to run- in one intersection, and somewhere the leadership of the region is firmly against point changes, arguing that they will break the existing infrastructure. Believe me, arguing with the management in all cases does not make sense, but from the point of view of logic, both options are suitable. Most often, in order to do nothing, they say: there is no money for the whole city, and the gradual introduction will break the logic of other traffic lights built over the years and lead to a collapse, such as if traffic flows go faster somewhere, in other places everything will stand up. We hear it all the time. But, in cities, every day there are some changes with the capacity: roads are blocked for repairs, new streets are built, streets are narrowed in winter, new traffic lights are installed, signs are changed… Nothing has yet led to a collapse, but on the contrary-it unloads traffic. Installing a single smart intersection is equivalent to building one more lane in all directions or putting a "stop prohibited" sign on the roadsides. Such a situation is easy to calculate, and rebuild the nearest traffic lights, as it has always been done for many decades. Talk of the end of the world is greatly exaggerated. Moreover, the neural network intersection sees the traffic flow and to the exit, so it can quickly rebuild, if it understands, that there is no point in allowing to pass transport in this particular direction.For the sake of information, it should be added that in a dozen years another type of Smart traffic lights will come - Communication. Even today, Volvo cars are equipped with equipment, that can transmit data about itself to all interested traffic control systems. You don't even need cameras. But for this it is necessary, that all the old models are rusted. Although it is possible, that such devices will become mandatory - as well as airplane detector, which were introduced after one of the air crashes. Of course, useful and life-saving technologies may not reach Russia for a long time. =========== Источник: habr.com =========== Похожие новости:
IT-инфраструктура ), #_big_data |
|
Вы не можете начинать темы
Вы не можете отвечать на сообщения
Вы не можете редактировать свои сообщения
Вы не можете удалять свои сообщения
Вы не можете голосовать в опросах
Вы не можете прикреплять файлы к сообщениям
Вы не можете скачивать файлы
Вы не можете отвечать на сообщения
Вы не можете редактировать свои сообщения
Вы не можете удалять свои сообщения
Вы не можете голосовать в опросах
Вы не можете прикреплять файлы к сообщениям
Вы не можете скачивать файлы
Текущее время: 22-Ноя 11:01
Часовой пояс: UTC + 5