Google head of unmanned vehicles: we build driver-Google, automatic driving, drone-IT information

Note: a duplicate of the original has been replaced.

Google is pioneer and leader in unmanned vehicle technology, as early as in 2009, the search giant has begun free vehicle research and development. But recently Google unmanned vehicle project director Dmitri Dolgov-explain ourselves to the Re/code robot car, says he did not unmanned vehicles, but drivers of a manufacturing business of Superman.

Dolgov's view is that driving is a social activity. In other words, unmanned vehicles must not only be able to detect and recognize people and other objects, to understand each other's behavior, and can interact with and react to.

To that end, Google's self-driving has developed a software that can detect and understand the gesture signals such as (of course, such gestures are pedestrians and the unity of human drivers on the road using symbolic communication), and then respond. For example cyclists left hand fisting thumbs up can be interpreted as intended to turn left (see Google for unmanned vehicle can be identified hundreds of bikes), unmanned vehicles will slow down to avoid.

For Google, it is the key to the sensor and the co-ordination of software, has little relationship with the manufacturer or model of the car.

Dolgov said:

Car will not matter. Because we build is a driver. We used Toyota Prius, Lexus and did his own prototype. And now we are in the new platform for cooperation with Fiat, Chrysler. As long as the software is good, what kind of car is the same. It's like you go into the other car. For example you go to rent a car--this may be bigger and drive with your own vehicle is a little different--you might want to take a moment to get used to, but the core tasks are not the same.

But Google no different is that it combines with a human driver maps and long Cheng Leida, at least within 100 meters of the conventional radar to see things also can know in advance where there must be detected before it turns and traffic lights and road.

The second difference is between unmanned vehicles will communicate through the cloud. This means that as long as no one learned anything new in the car driving skills, other unmanned vehicles also can learn right away.

Of course, this will result in unmanned vehicles being black may. Despite the start-up companies in tackling the safety of unmanned vehicles, Google also has a security solution. That is isolated communication channels as well as for the electronic control of the channel. Google will limit the amount of information that can be shared, and delineated a security zone, makes it impossible for external intruders control of local driving.

In fact, Google tried to make not so much drivers, rather than doing things, sensors and artificial intelligence technologies enable under-aware computing, allowing all its artifacts, including cars have the ability to act with wisdom, just as biological (creature).

Google无人车负责人：我们造的是司机 - 谷歌,自动驾驶,无人驾驶 - IT资讯

注：原文重复，现已替换。

Google无疑是无人车技术的先驱和领先者，早在2009年搜索巨头就已经开始了无人车的研发。但是近日Google无人车项目负责人Dmitri Dolgov在向Re/code介绍自己的无人车时，却说自己做的不是无人车，而是一门制造司机超人的业务。

Dolgov的观点是，驾驶是一项社会活动。也就是说，无人车不仅要能检测和识别出人等其他对象，还要能理解对方的行为，并且还要能与之交互以及做出反应。

为此，Google的自动驾驶软件已经发展到可以检测和理解手势之类的信号（当然这种手势应该是行人和人类司机在道路上共同使用的统一符号性沟通方式），然后做出响应。比方说自行车手伸出左手握拳大拇指向上就可以被解读为打算左拐弯（参见Google的无人车可同时识别上百辆自行车），无人车就会减速避让。

对于Google来说，做成这件事的关键是传感器和软件的配合，与汽车的制造商或者车型关系不大。

Dolgov说：

车的关系不大。因为我们造的是司机。我们以前用丰田普锐斯、雷克萨斯和自己的原型车都做过。而现在我们则在跟菲亚特克莱斯勒在新平台合作。只要软件处理好，什么车都一样。这就好比你进入另一辆车一样。比如你去租一辆车——这车可能会大一点，开起来跟你自己的车有点不一样——你可能要花点时间去习惯，但核心的任务不变。

不过Google无人车跟人类司机不同的是它要结合地图和长程雷达、常规雷达去看清至少100米范围内的东西，同时还必须在检测之前就能事先知道哪里有转弯、交通灯和道路。

第二个不同是无人车相互之间还会通过云进行沟通。这意味着只要一台无人车学会了什么新的驾驶技能，其他的无人车也马上就能学会。

当然，这会导致无人车被黑的可能。尽管已经有初创企业在攻关无人车的安全问题，Google自己也有保护安全的解决方案。那就是隔离相互沟通的通道以及用于电子控制的通道。Google会限制可共享的信息量，并且划出了一个安全区，使得外部入侵者不可能控制本地的驾驶行为。

其实，与其说Google试着造司机，不如说是做物联网、传感器及人工智能技术使能下的感知计算，从而让包括汽车在内的一切人造物具备行动能力与智慧，就像生物（creature）一样。