How do self-driving cars “see”? - Sajan Saini

翻译人员: Carol Wang 校对人员: Ziyao Wang

It’s late, pitch dark, and a self-driving car winds down a narrow country road.

天已晚，黑夜中一辆自动驾驶汽车 沿狭窄的乡村公路蜿蜒行驶而来。

Suddenly, three hazards appear at the same time.

突然间，同时出现了三个危险警示。

What happens next?

接下来会发生什么呢？

Before it can navigate this onslaught of obstacles,

在驾驶通过突然出现的障碍物前，

the car has to detect them—

汽车必须先发现障碍物——

gleaning enough information about their size, shape, and position,

收集有关大小、形状和位置信息，

so that its control algorithms can plot the safest course.

以便用算法规划出最安全的路线。

With no human at the wheel,

因为无人驾驶，

the car needs smart eyes, sensors that’ll resolve these details—

汽车需要智能眼， 也就是解决问题的探测器——

no matter the environment, weather, or how dark it is—

无论周围环境、天气如何， 或者天有多黑——

all in a split-second.

问题要在眨眼间解决。

That’s a tall order, but there’s a solution that partners two things:

要求虽高，但有解决方案， 该方案由两样东西组成：

a special kind of laser-based probe called LIDAR,

特殊激光探测技术 LIDAR

and a miniature version of the communications technology

和集成光电技术，

that keeps the internet humming, called integrated photonics.

即实时联网的微通信技术。

To understand LIDAR, it helps to start with a related technology— radar.

要了解光达 LIDAR，需要先了解 与此相关的技术 ——雷达。

In aviation,

在航空领域，

radar antennas launch pulses of radio or microwaves at planes

雷达天线向飞机 发射无线电脉冲或微波，

to learn their locations by timing how long the beams take to bounce back.

通过计算波束返回时长 来确定飞机位置。

That’s a limited way of seeing, though,

但这一种观察方式很有限，

because the large beam-size can’t visualize fine details.

因为大光束脉冲或微波 无法显示微小的细节。

In contrast, a self-driving car’s LIDAR system,

相反，自动驾驶汽车的 LIDAR 系统

which stands for Light Detection and Ranging,

使用窄带不可见红外激光，

uses a narrow invisible infrared laser.

LIDAR 是光探测和测距的缩写。

It can image features as small as the button on a pedestrian’s shirt

它可以将极其微小物件成像，

如街对面行人衬衫上的扣子。

across the street.

But how do we determine the shape, or depth, of these features?

但是，我们如何确定 这些物件的形状或距离呢？

LIDAR fires a train of super-short laser pulses to give depth resolution.

LIDAR 发射一系列超短 激光脉冲去测距离。

Take the moose on the country road.

以乡间小路上的驼鹿为例，

As the car drives by, one LIDAR pulse scatters off the base of its antlers,

汽车驶过时，一个 LIDAR 脉冲 遇到鹿角的根部后散开，

while the next may travel to the tip of one antler before bouncing back.

它反弹回来前，下一个脉冲 可能已到达一个鹿角的顶端。

Measuring how much longer the second pulse takes to return

测量这两个脉冲返回的时长差，

provides data about the antler’s shape.

可得到有关鹿角形状的数据。

With a lot of short pulses, a LIDAR system quickly renders a detailed profile.

通过发射大量短脉冲，LIDAR 系统可快速得出物体详细轮廓。

The most obvious way to create a pulse of light is to switch a laser on and off.

打开再关上激光器是 产生光脉冲最简便的方法，

But this makes a laser unstable and affects the precise timing of its pulses,

但会造成激光束不稳定， 并影响发射脉冲的精确频率，

which limits depth resolution.

影响距离测量准确性。

Better to leave it on,

更好的办法是让激光器开着，

and use something else to periodically block the light reliably and rapidly.

用其他东西定期、快速地阻挡光线。

That’s where integrated photonics come in.

这就是集成光子技术。

The digital data of the internet

互联网的数字数据

is carried by precision-timed pulses of light,

是由精确定时的光脉冲承载的，

some as short as a hundred picoseconds.

有的脉冲短至一百皮秒。

One way to create these pulses is with a Mach-Zehnder modulator.

一种产生光脉冲的方法是 使用马赫-曾德尔干涉仪，

This device takes advantage of a particular wave property,

该设备利用特定的波特性，

called interference.

称为干扰特性。

Imagine dropping pebbles into a pond:

想一下将一些鹅卵石 扔进池塘的情景：

as the ripples spread and overlap, a pattern forms.

涟漪扩散和交叠 构成了一种花纹，

In some places, wave peaks add up to become very large;

某些地方的波峰 叠加，变得非常大;

in other places, they completely cancel out.

而其他地方，则完全抵消了。

The Mach-Zehnder modulator does something similar.

马赫-曾德尔 干涉仪的原理与此类似。

It splits waves of light along two parallel arms and eventually rejoins them.

先将光波沿 2 个平行臂 分为 2 束，然后合二为一。

If the light is slowed down and delayed in one arm,

如果一束光减慢而延迟，

the waves recombine out of sync and cancel, blocking the light.

因为 2 束光不同步，合并后的 抵消现象就阻挡了光线。

By toggling this delay in one arm,

通过切换一束光的延迟，

the modulator acts like an on/off switch, emitting pulses of light.

干扰器就像一个开合的开关， 发射出光脉冲。

A light pulse lasting a hundred picoseconds

持续一百皮秒的光脉冲

leads to a depth resolution of a few centimeters,

可以探测到小至几厘米厚的物体，

but tomorrow’s cars will need to see better than that.

但未来的汽车需要更高的分辨率。

By pairing the modulator with a super- sensitive, fast-acting light detector,

将干扰器与超灵敏、 反应快的光探测器配对，

the resolution can be refined to a millimeter.

可将分辨率提高到毫米级。

That’s more than a hundred times better

比我们以正常视力看街对面的物体，

than what we can make out with 20/20 vision, from across a street.

要好一百倍以上，

The first generation of automobile LIDAR has relied on complex spinning assemblies

第一代自动驾驶汽车的 LIDAR

依赖车顶或发动机盖上的 复杂旋转组件进行扫描。

that scan from rooftops or hoods.

With integrated photonics,

借助集成光子技术，

modulators and detectors are being shrunk to less than a tenth of a millimeter,

干扰器和探测器可缩小至 不到十分之一毫米，

and packed into tiny chips that’ll one day fit inside a car’s lights.

装在小巧的芯片中， 将来可以放在车灯里。

These chips will also include a clever variation on the modulator

这些芯片还将包括 干扰器智能调节器，

to help do away with moving parts and scan at rapid speeds.

可消除移动物体，并快速扫描。

By slowing the light in a modulator arm only a tiny bit,

通过减慢干扰器一个平行臂发出的光，

this additional device will act more like a dimmer than an on/off switch.

这个额外设备的作用不像开关， 更像一个调光器，

If an array of many such arms, each with a tiny controlled delay,

如果许多带微型控制延时的 平行臂并列排放的话，

is stacked in parallel, something novel can be designed:

就设计出了新的特性：

a steerable laser beam.

可操纵的激光束。

From their new vantage,

有了这些新优势，

these smart eyes will probe and see more thoroughly

这些智能眼的探测和观察

than anything nature could’ve imagined—

比能想到的任何自然的东西更彻底——

and help navigate any number of obstacles.

帮助导航通过任何数量的障碍物,

All without anyone breaking a sweat—

不费吹灰之力——

except for maybe one disoriented moose.

或许一头没有方向感的驼鹿除外。