【无人驾驶汽车像人类司机学习】

【无人驾驶汽车像人类司机学习】日前,汽车制造商Jaguar Land Rover、技术公司 Bosch 及交通研究实验室联合推出 Move-UK计划,旨在提高电脑识别风险的能力。他们在人类驾驶的汽车中安装传感器和自动驾驶系统,自动系统将在后台运行,其做出的判断将与人类的真实反映进行对比。这将帮助电脑更好地识别风险并作出最佳反应。

Self-driving cars to learn from human drivers: Project could help prevent accidents like the Tesla crash

Driverless cars are being taught to drive more like human motorists in an attempt to help them recognise and respond to risks on the road.

The technology is being developed in the hope of overcoming some of the problems that are thought to have led to the fatal crash involving a Tesla electric car with its autopilot feature enabled.

Scientists and road safety campaigners insist there is still significant development needed before motorists will be able to take their hands entirely off the steering wheel and let robots take over.

35F002F100000578-0-Driverless_cars_are_to_learn_from_human_drivers_about_how_to_spo-a-2_1467632953814

Joshua Brown, 40, a former Navy SEAL, died in May after his Tesla Model S collided with a truck on a freeway in Williston, Florida, after it failed to detect the white trailer against the sky.

An investigation into the crash is still ongoing by the US National Highway Traffic Safety Administration, but there are reports that Mr Brown had been watching a movie at the time.

Tesla insisted its Autopilot feature was not intended to take over total control of the vehicle and it made clear to motorists that they must stay alert and prepared to take over at any time.

However, the accident has led many to question the safety of putting autonomous and semi-autonomous vehicles on the roads with human drivers.

Research by the Transport Research Laboratory in Wokingham, Berkshire, has shown that drivers can easily become distracted once they believe a vehicle is capable of doing the driving itself.

But a new project known as Move-UK, conducted by car manufacturer Jaguar Land Rover, technology firm Bosch and the Transport Research Laboratory, is aimed at improving the way computers spot risks on the roads.

They are installing sensors and self-driving systems of a series of cars that will be driven by humans through the city streets and on public roads.

The autonomous system will operate in the background without controlling the vehicle and its decisions will be compared to the real life actions of the human driver.

This, engineers behind the project say, will allow them to develop better ways of spotting risks and teach the computers in which situations it is best to react.

微信截图_20160705101751 360截图20160705080531388

Richard Cuerden, chief scientist engineering and technology at the Transport Research Laboratory, told MailOnline: 'All sorts of things can make driving on the roads unpredictable.

'The weather, different lighting conditions and traffic can all change how we drive and react.

'We want to use this project to learn from the data how we can predict risk. There will be times when a computer identifies risk and the drive made no action and everything turned out ok.

'We will want to explore why this might be and what actions are appropriate. There may well be unfortunate events were the driver does not do anything and the computer would have.'

Since Tesla released its Autopilot feature there have been numerous videos that have appeared online showing motorists driving along roads without their hands on the steering wheel.

Tesla claims its cars have logged up 130 million miles using its Autopilot feature with just one known fatality.

AA president Edmund King claimed the accident in the US is a reminder that 'driverless cars aren't fool proof in the real world'.

He said: 'We need more research into the interactions between driverless cars and driver-driven vehicles before we allow all drivers to take their hands off the wheel.'

The Autopilot feature is what is known as a Level 2 autonomous technology, where the car can take over basic functions like staying in a lane, and braking if it gets too close to a vehicle in front.

The next generation of autonomous vehicles are aimed at being Level 4, where motorists will be able to let the car do all of the driving.

Google apparently chose to develop its self-driving cars at the Level 4 of autonomy after finding motorists in its early test vehicles quickly became distracted, rummaging in bags, doing make up and taking their eyes of the road for long periods.

It decided that full autonomy was the only way to reduce the risk of accidents.

Experts say that learning from accidents involving autonomous features will be essential to improving the technology.

Professor Nick Reed, technical leader of the Greenwich Automated Transport Environment project that is testing the use of self-driving cars in London for the UK Government and who has studied how humans interact with autonomous vehicles, warned there are likely to be more accidents involving such vehicles in the future.

He said: 'There is a tendency for people to over trust autonomous technology. When they see it behaving in a way they expect – like making turns and braking, people's attention can quite easily lapse as they feel comfortable.

35F0083800000578-3673554-Jaguar_Land_Rover_is_developing_technology_that_it_hopes_will_le-a-5_1467633900459

'Although we may see more of these kind of accidents in the future, the technology will potentially prevent more of the frequent kind of accidents like riding into the back of the car in front.

'We will see a different kind of accident but fewer accidents over all.'

03431D84000003E8-3673554-image-a-3_1467633644583

However, it is hoped the Move-UK project will allow engineers to fine tune the technology to make fully self-driving cars safer than human drivers.

Dr Wolfgang Epple, Director of Research and Technology, Jaguar Land Rover, said: 'To successfully introduce autonomous cars, we actually need to focus more on the driver than ever before.

'Understanding how drivers react to a range of very dynamic and random situations in the real world is essential if we want drivers to embrace autonomous cars in the future.'

http://www.dailymail.co.uk/sciencetech/article-3673554/Self-driving-cars-learn-human-drivers-Project-teach-autonomous-vehicles-spot-risks-help-prevent-accidents-like-Tesla-crash.html


Comments are closed.



无觅相关文章插件