Future cars: Cloud computing predicts driving behavior against car crashes

Traffic jam: As long as there is a car, we have traffic jam. But a new generation of smart car technology is helping drivers and cars better travel, prevent traffic jams, avoid car crashes, and relieve 5 o'clock in the afternoon traffic rage.

This screenshot shows the Smarter Traveler application of American International Business machines. It uses the smartphone's GPS system to track drivers ' habits and use a road sensor network to check traffic in specific areas. The company has developed a calculation of the future traffic forecast results based on traffic flow, the application can send the driver to the blocked section before sending them short message or mail

Sina Science and technology news Beijing time November 29, this July, the U.S. state of Nevada Parliament passed a resolution calling for the development of unmanned vehicle traffic regulations, this is a very far-sighted move. Many vehicles have been involved in driving activities, and in the past, driving was purely controlled by people, such as parking and braking control, but now, some vehicles have been able to do automatic parking and braking, of course, they can not automatically navigate, make their own decisions and driving. But in some ways, the autonomous vehicles have begun to integrate into our lives. Using relatively simple software to adjust existing hardware, major U.S. and European carmakers are making cars smarter and greener through hybrid engines.

Cars are connected to each other and "cloud" together, they can make their own decisions, so the future of driving will largely be done by the car itself, without human intervention. Algorithms and analysis will predict the driver's behavior, and then give the most idealized route for the driver to choose from. Radar-equipped private cars will be able to sense the environment around them, rail traffic, car and car networks, can reduce traffic jams, prevent accidents, improve fuel economy. "A lot of the interactions we have now can help drivers drive better, but we're trying to help cars do these things and make them smarter," said Ryan Maikhi, an engineer at Ford Research and Innovation Center in Durpon, Michigan. When you drive a car and connect it to the cloud, there are many possibilities. ”

You can predict where you're going.

If you don't use public transport, you may be driving by yourself, commuting to work or school on a specific route every day, such as going from home to the coffee shop in the morning and then to the office, leaving the office at night and probably going to the store to go shopping before you go home.

Ford wants to use your predictability to guess where you're going when you start your car. The researchers are feeding the driving history back to Google's software server "Predictive API", which uses machine learning algorithms to generate predictive behavior patterns, especially drivers ' prediction. Maikhi is studying the model, he said, "The question we ask the model is, ' where is the next person going?" ' This model will answer that it's Wednesday 5 o'clock in the morning, so you're probably going home, and it will feed back to the car. ”

The current network-connected system is mainly used to record the day of the car, at what time, and where. The algorithms will work out a range of possible routes and departure and arrival times. According to Maikhi, cars will automatically change power consumption and replace petrol with batteries, which will be useful for plug-in hybrids, depending on the itinerary. Ford has experimented with an escape SUV plug-in hybrid. In the future, the Government may strengthen low emissions areas near schools or parks, or build special routes for battery-powered vehicles. Maikhi on the way home from the unit, often by the way of the son who is practicing football. If his car knew about his habit, he could save some of the battery energy in a day's operation by knowing where the low emission area was.

"It will ensure that I save some energy when I go there and I can drive cars with electricity," Maikhi said. If we know how people will use cars, we can try to optimize their performance. "But all the routines will be broken, and if you want to eat a burrito in Wednesday and you want to eat a mince sandwich in Thursday, what about the car?" Lunch time it will not be clear where you are going. Because of the randomness and the flexibility of the driver, the person who drives the car can be added and adjusted at any time if it is installed in a mass-market model.

Cars predict future traffic

Once the car is clear about the driver's habits, it can combine other data in the cloud to provide better advice to drivers. If traffic forecasters have enough data and they have good models to explain it, they will tell you at lunchtime when they get home 5 o'clock in the afternoon.

IBM is using a smartphone application to do this. The company said the app knew the driver's routine, predicted traffic for the next 30 minutes or more, and sent the results to the drivers before they reached the congested section. The Blue Giant (IBM nickname) has tested the system in the San Francisco Bay Area, according to John Dai, an international commercial machinery company's Intelligent Transportation Project manager.

The app uses a mobile phone's GPS to track the driver's activities at different times of the day, knowing her or his path. It is also connected to 700 road sensors previously installed by the California State Transport Authority. These sensors track the speed of cars and the flow of traffic, and calculate how few cars pass each minute. They collect data every 30 seconds, but IBM feeds the data back to the algorithm once every 5 minutes. Dai said that the algorithm based on the relationship between these data, the development of driving patterns. "This tool is good at discovering and understanding deceleration signals," he said. It will focus on a particular area, especially the switching position, and when the traffic slows down, you are 83% more likely to get stuck in traffic jams. ”

Using these signals, the system will analyze the real-time data and obtain an updated traffic situation model, which is updated every 5 minutes. While existing forecasting systems are based on their predictions of current traffic conditions, if traffic conditions are predicted without any change, the algorithms can recognize and consider the ripple effects of a single action. Users can view "driving history" through the computer, and can add or remove routes to control the amount of information stored. Dai even imagined that the app could offer coupons to people who often drive. "It would be very useful if you had the ability to identify people who were driving past a coffee shop," he said. However, use this information carefully, regardless of whether the user wants to share the data, should be well managed. But it does have that ability right now. ”

But the main use of the app is to help drivers avoid traffic jams, he says. It will provide an optimized route, or even check the network's available schedules. If a driver often leaves home to work at 8:30 A.M., he may receive traffic information at 8 o'clock in the morning, "your usual route today doesn't look very good, but the train leaves at 8:30 A.M., and 48 parking spaces are available at the stop and transfer point." Dai says the system works in any place where road sensors and drivers have GPS navigation phones. "Take a look at the news before you go to any city, and you'll see some hot pictures, and some like the weatherman are talking about traffic," he said. ”

Predicting the movements of pedestrians and other drivers

Your car now knows what you're going to do next, and the traffic, and it needs to know what to do next to the 20-foot (6.1-meter) vehicle ahead. The network between cars and cars and advanced control algorithms can ensure that there will be no unexpected events, so there will be no future accidents. They also improve engine efficiency by preventing cars from walking, stopping traffic, and reducing emissions.

Recently, the U.S. Department of Transportation announced a pilot program that drivers can use to test the capabilities of future interconnected cars (connected car). These systems allow for "communication" between cars, but drivers first have to use them. If we can't adapt it well, the car will solve the problem quickly. MIT researchers are working on a new algorithm that combines human behavior patterns to warn drivers of possible car crashes and controls to avoid accidents.

The system aims to shrink the driving process into two steps: braking and acceleration. According to the Massachusetts Institute of Information, the driver at a certain point in time will produce a lot of results, and then he may go to many places. Domidi-Vikiro, a professor of mechanical engineering at the school, also combines the ability to predict human behavior, such as when a driver slows down or accelerates at a crossroads. The results of this system can determine at what point the vehicle at the crossroads is dangerous, the vehicle with this budget rules will try to predict what the other cars will do next, and refer to the traffic lights and car sensors information, as far as possible to avoid accidents.

Since there is no need to drive in the future, algorithms must take into account the variable factors of vehicle behavior and consider the chain reaction caused by an action. For example, if a car slows down in order to avoid a crash, other cars must also change their speed. The distributed control system controls the acceleration, braking, two-lane transformation, and road exit for all cars in a particular group.

Researchers at Carnegie Mellon University have produced a simulation tool that can prove the safety of the system, and even multiple vehicles for complex tasks. Associate professor of computer science A team of researchers, led by Andre-Phillips, initially experimented with two vehicles on a one-way street. They then added more vehicles to confirm how many vehicles the system would work well and add more lanes to verify that the driveway had no effect on it. Finally, no matter how many cars and how many lanes, the system can effectively prevent the crash. Future simulations will take into account variables such as detours.

It will take a little longer for automatic fleet control systems to become reality. Instead of using advanced algorithms to reduce congestion and prevent traffic accidents at the same time, European researchers have done so only by allowing vehicles to "talk" to each other. Such systems can reduce emissions and improve engine efficiency. Volvo, the Swedish carmaker, has been experimenting with "road trains," which involve car escorts by professional drivers. The car can be hung on a train, connected to it by a wireless line, and the adaptive fixed-speed cruise system must match the speed of the road train. The sensors in the car will be used to ensure a safe distance between everyone. Earlier this year, Volvo experimented with road trains through a car and a semitrailer. It is designed to save fuel, reduce road congestion and ensure driving safety. The carmaker hopes to start using the new technology in 2020.

In June this year, German researchers found that if 5 vehicles in each of the 1000 vehicles communicated with each other, they could reduce road congestion. The sensors in the car can collect data and exchange information with other vehicles through the local wireless network and transfer these to the Traffic command center. Preliminary data collected in a one-year trial show that at least 5 vehicles can have a significant impact on traffic congestion. The experiment is part of the EU Diamant project.

The good news for carmakers and consumers is that most of these technologies can be implemented with simple coding and algorithmic changes, and that cars need not be redesigned for improvements. By updating data like this, driving will be safer and less congested in the next few years. (Emperor)

(Responsible editor: The good of the Legacy)