Photo: Look no eyes! Sandstorm, a self-driving car from the 2005 DARPA Grand Challenge, uses sensors of different kinds to "see" where it's going, including short- and long-range LIDAR. The long-range LIDAR scanner is the round, white object on top of the roof; it can "see" a distance of 150m (450ft); the short-range LIDAR units are mounted on the front of the car near the radiator and to the left of the windshield area. Photo courtesy of Dan Homerick published on Flickr under a Creative Commons Licence .
That's why autonomous robots (ones that control themselves) and self-driving cars often prefer to look at the world a different way, using LIDAR systems instead of cameras. Where a camera-based eye snaps an instant 2D photo of a scene that has to be processed and interpreted to find out what it's looking at, LIDAR makes millions of measurements of depth information in all directions simultaneously—and it's often quicker and easier to turn that data into a map you can use for navigation, in real-time.
Look around you. What you see is a 3D color map of your immediate environment that your brain has built (mostly in real-time) using the light rays soaked up by your eyes. If you were a robot with a couple of digital cameras stuck on your head, you could build yourself a map of a room in much the same way, but it wouldn't be anything like as informative and useful. You wouldn't necessarily know that one object was nearer than another or that a growing black blob in the middle of the room was a cat creeping toward you. As a human, you know these things because your brain processes visual information using a lifetime of experience of what a growing black blob actually means. But robots don't have the same encyclopedic life experience to draw on, which means they're at a natural disadvantage when it comes to "seeing" the world.
Artwork: The basic idea of navigational LIDAR: the self-driving car (blue) "sees" by bouncing a spinning laser beam (orange lines) off obstacles and detecting its reflections (green lines). The time it takes for the beam to return tells us how far each obstacle is from the car. In this way, LIDAR creates a 3D map of the dynamic environment around the car much more quickly than the car itself is driving.
Don't know where you're going... don't know how to get there? Then you'll need a map. But what if there is no map? Hmmm.... okay, then what if part of your brain could rush ahead of you, quickly sketch a map, and feed it back to the rest of your brain to help you find your way? It sounds completely bonkers, but it's exactly how self-driving cars work—using a neat 3D map-making technology called LIDAR (which stands for LIght Detection And Ranging). As the name suggests, LIDAR works a little bit like radar (radio-wave navigation used by ships and planes) and sonar (underwater detection using sound, mainly used by submarines ), though it has quite different applications, from factory-floor robots and driverless taxis to coastal mapping and measuring deforestation. How exactly does it work? Let's take a closer look!
Take a look at a factory-floor robot or a self-driving car. What's that weird, whirling can thing sitting on top like a helmet? That's the LIDAR: it's spinning round, firing invisible laser beams in all directions, catching the reflections, and measuring how long the beams take to return so it can figure out what obstacles are nearby and how far away they are. So the basic concept of LIDAR is exactly the same as radar and sonar. With radar, you might have a jet plane firing out a beam of coded radio waves and listening for a return beam reflected off some nearby object (another plane about to crash into you); it uses the time taken for the beam to return to figure out how far away the object is. With sonar, you do the same thing underwater, only using sound waves (because ordinary light and radio waves don't travel through water very far). In everyday, on-land situations—driving down the street or navigating through a building—reflected laser light turns out to be a better source of information than either radio waves or sound, and that's why LIDAR has become so popular: it's simple, reliable, and relatively low-cost, if still very expensive for amateur or hobbyist use (currently, we're talking thousands of dollars).
So you can use LIDAR data to build a real-time map of the streets through which a self-driving car is trying to navigate or the factory a robot has to trundle through, but you can also it in other ways. Long before self-driving cars became such a hot topic, geographers and atmospheric scientists were using LIDAR in a more "passive" way, to draw detailed aerial maps of Earth or the atmosphere. (Climate scientists proposed studying the weather with laser-radar as far back as the mid-1960s.) For these sorts of applications, instead of using a spinning LIDAR, you mount a LIDAR unit underneath a plane or a helicopter and have it scanning across the ground as the plane flies across a precise trajectory. While visible light isn't much use for scanning underwater, it is possible to use blue-green laser light to make LIDAR scans of the seabed, for example.
Artwork: LIDAR is now probably best known for its use in robots and self-driving cars, but it has many other applications. High spectral resolution LIDAR (HSRL), often operated from scanning airplanes like this, is used to study Earth's atmosphere and oceans. It works by sending out LIDAR signals, then examining the spectrum of the radiation that's "backscattered" (which means roughly rather than exactly reflected) by molecules and aerosols in the atmosphere.
LIDAR data can be used by itself or combined with data gathered in other ways. In the case of an aerial map, LIDAR systems typically also use GPS (satellite navigation); in self-driving cars, LIDAR tends to be used alongside GPS, onboard sensors (like accelerometers or speedometers), inertial guidance systems and gyrocompasses, and navigational data from stored maps (think Google Street View). What you end up with is called a "point cloud": a three-dimensional array of LIDAR measurements related to specific GPS coordinates. For a moving sensor, like the one on a self-driving car, you end up with millions of data points stretching as far as 60m (200ft) away from you in all directions, accurate to just a few centimeters.
Photo: Mapping the Moon. NASA engineers and astronauts test K10 lunar rover scanning units at Moses Lake dunes. The rovers have ground-penetrating radar and LIDAR scanners that can make 3D maps of the Moon's terrain, both above and below ground. Photo courtesy of NASA on the Commons.
To make a LIDAR map, you need a laser and something that picks up its reflected light; something to move your laser beam and make it scan all around you; and typically also a GPS receiver so you can figure out where you are and which bit of the world your LIDAR data applies to.
Photo: What do you need for LIDAR scanning? Something like this truck, set up by researchers at the US Department of Agriculture to fire green LIDAR beams into the air from a scanner on the roof. LIDAR is widely used for mapping the world and studying its atmosphere. USDA researchers have been using it for 20 years to study how agriculture and farming affect the atmosphere—everything from water uptake by growing trees to the air pollution made by intensive animal farming. Photo by Peggy Greb courtesy of USDA Agricultural Research Service.
Will any old laser do? One of those big, buzzing lasers like Goldfinger wanted to use to cut James Bond in half? No! Typically, we're talking about a semiconductor diode laser, more like the ones you'd find in a laser printer or CD player only more powerful. Instead of firing out visible light (which has wavelengths of around 400–700 nanometers), a self-driving car would use a LIDAR with an invisible, near-infrared laser (around 900–1100 nanometers). Underwater LIDAR scanners use green laser light with shorter wavelengths (around 530 nanometers) in the middle of the visible range. There's obviously some danger to people's eyes when you start firing infrared laser beams all over the place—and more danger with cars hurtling down the street than with airplanes scanning distant rainforests from the sky. Generally speaking, the further a LIDAR laser needs to penetrate safely, the higher the wavelength it will use—because light of longer wavelength has a shorter frequency and lower energy. The latest self-driving car lasers are using laser wavelengths of 1550 nanometers to scan up to 200 meters ahead, compared to just 30–40 meters for a high-powered laser working at 905 nanometers.
The photodetector in a LIDAR system is a kind of photoelectric cell made of silicon or gallium arsenide that's designed with maximum sensitivity for whatever light wavelength the laser is using. Different types of detectors are used according to the kind of range over which the LIDAR system is operating. Short-range LIDAR systems typically use simple silicon photodiodes. Long-range systems use what are called avalanche photodiodes (APDs). These work a bit like Geiger counter radiation detectors, turning a single incoming photon of light into a measurable avalanche of electrons (an electric current that can be measured), so much lower light levels can be detected. Many APDs can be built into a single chip to create a kind of checkerboard of detectors called a multi-pixel photon counter (MPPC).
Spinning a laser at high speed sounds like a bit of an engineering nightmare—all those tangled wires and vibrating metal cases—but, fortunately, we can get by without doing that. All we have to do in a LIDAR system is scan the beam, not the laser itself, and for that we can just use a rapidly rotating mirror. Modern LIDAR systems use microscopic moving mirrors based on MEMS (MicroElectroMechanical Systems) technology, mounted on microchips and similar to the ones you find in digital projectors; other use bigger mirrors roughly the size of coins.
Animation: 1) In theory, LIDAR lasers are scanned by firing them off a fixed mirror (top) and a rapidly rotating one (bottom).
Artwork: 2) In practice, it's not quite so clunky, and modern LIDARs tend to use very small microscopic mirrors based on MEMS technology. Each tiny mirror segment (pink) tilts on a hinge (green), attracted by the electrically charged plates below (blue). Artwork from US Patent 4,710,732: Spatial light modulator and method by Larry Hornbeck, Texas Instruments, 1987, courtesy of US Patent and Trademark Office.
Though robots and self-driving cars are the kinds of LIDAR applications you're likely to read about in the technical press, the most common applications to date are in geographical and atmospheric mapping. People like the USGS (US Geological Survey), NOAA (National Oceanographic and Atmospheric Administration), and NASA have been using LIDAR for make maps of Earth and space for decades. Climate scientists use it to probe the composition of the atmosphere and study things like clouds, aerosols, and global warming; oceanographers use it to track coastal erosion; and botanists are flying up in airplanes right now using LIDAR to measure the ever-changing patterns of Earth's forests. LIDAR's been used for everything from monitoring spider populations and studying threatened butterflies to modeling oak trees and monitoring beach damage after storms. It's even been used to reveal traces of long-lost civilizations buried under forests.
Photo: SeaHunter, a self-guided prototype aerial vehicle, uses LIDAR to navigate. Photo by Grant P. Ammon courtesy of US Navy.
We can also use LIDAR to study the gas composition of the atmosphere. Different gases absorb different wavelengths of light by different amounts, so we can study the gases in a particular location remotely by firing two different wavelength laser beams into it from a plane or a helicopter and comparing how much of each wavelength is absorbed or reflected. This system, which is called Differential Absorption LIDAR (DIAL), can be used for everything from detecting leakages from gas pipelines to measuring air pollution.
One of the most common uses of LIDAR is in police speed guns. Though we typically think of them as radar guns (and static highway speed cameras, such as Gatsos, do use radar), handheld guns are much more likely to use 905-nanometer LIDAR lasers, which are cheap, safe, and very effective.