Self-driving cars will bristle with sensors

Car computers will use many sources of data -- lasers, radar, stereo cameras, even windshield wiper rain detectors -- to figure out what's around them. And none of the sensors will ever get drowsy.

Stephen Shankland principal writer
Stephen Shankland has been a reporter at CNET since 1998 and writes about processors, digital photography, AI, quantum computing, computer science, materials science, supercomputers, drones, browsers, 3D printing, USB, and new computing technology in general. He has a soft spot in his heart for standards groups and I/O interfaces. His first big scoop was about radioactive cat poop.
Expertise processors, semiconductors, web browsers, quantum computing, supercomputers, AI, 3D printing, drones, computer science, physics, programming, materials science, USB, UWB, Android, digital photography, science Credentials
  • I've been covering the technology industry for 24 years and was a science writer for five years before that. I've got deep expertise in microprocessors, digital photography, computer hardware and software, internet standards, web technology, and other dee
Stephen Shankland
5 min read
Google's computer systems analyze signals from a laser scanner to identify cars, trucks, pedestrians, and other roadway features.
Google's computer systems analyze signals from a laser scanner to identify cars, trucks, pedestrians, and other roadway features. Google

Editors' note: Be sure to catch the other stories in this package: on how Google's robo-cars mean the end of driving as we know it, on real-world experiments with platoons of connected cars, and on smart transport grids.

Self-driving cars are analogous to another autonomous entity, the human being: the car's computer serves the controlling role that our brain does, while its sensors do what our own senses do.

But in the computing-infused future of automobiles, the parallel only goes so far when it comes to sensors. Not only are cars able to detect radio and light waves that humans can't, they also never get drowsy and they can see in all directions at the same time.

The shift to self-driving cars will turn vehicles into roaming sensoriums connected to heavy-duty computers. Google's self-driving cars, for instance, use quad-core PCs that each second process 1.3 million laser measurements and make 20 driving decisions.

The sensor transition has already begun with the arrival of very small computer systems called microcontrollers that are in charge of cars' traction control systems, engines, airbags, and antilock brakes.

Cars and computers: A look at the future of autos (pictures)

See all photos

But more sensors and data processing are required as cars take more steps down the path of autonomy. Cars are gaining the ability to parallel park on their own, for example, and to apply the brakes to prevent or ameliorate a rear-ender when the car in front slows abruptly. Also arriving is the ability to take over the acceleration, braking, and steering for people stuck in start-and-stop traffic.

These sorts of features are possible because of electronics deeply embedded in car control systems. Steering wheels no longer directly turn gears, brake pedals no longer directly push hydraulic fluid, and accelerator pedals no longer directly pull throttle cables. Instead, computer microcontrollers interpret the controls' mechanical changes and then take the appropriate actions. This "drive by wire" technology means electronics already are are an essential part of driving. That change opened the door for computers to supplement human judgment.

To give computers control over driving, more sensors are needed. Gyroscopes and accelerometers tell the car when anti-rollover maneuvers are appropriate. Gauges wirelessly beam tire pressure to the car's control system. Ultrasonic sound systems warn of lurking obstacles when parking. Ambient light sensors turn headlights on and off. Moisture detectors turn on windshield wipers.

That's today's leading-edge technology. Tomorrow, a host of new sensors will become common. Radar will detect nearby cars for collision avoidance. Cameras will keep track of road lanes and read traffic signs.

Bosch has been testing self-driving car technology in Silicon Valley.
Bosch has been testing self-driving car technology in Silicon Valley. Declan McCullagh/CNET

And perhaps most powerfully, lidar (light detection and ranging) laser sensors could generate a constantly updated 3D map of the car's vicinity. High-end lidar detectors are very expensive right now, but they've already made a showing, for example in Volvo's City Safety system that detects when cars ahead are stopping. Self-driving car systems from Bosch and Google are crowned by lidar scanners.

Lidar is great, but it has problems. Building that capability into bumpers is convenient but exposes the requisite gadgetry to damage, Alberto Broggi, a professor of computer engineering at the University of Parma in Italy who was named an IEEE senior member for his autonomous vehicle work, said the units he uses in his research cost about $20,000 apiece.

"Every year I talk to the laser guys," Broggi said. "They always tell me next year we will see lasers drop to 300 [euros, or $400] each. Next year, next year, Next year."

DOT V2V safety device
The Department of Transportation is testing acceptance of vehicle-to-vehicle (V2V) communication, a technology that could provide either human drivers or car computers with very important data. DOT

Sensors are getting cheaper as they move to microprocessor-derived manufacturing methods. "There is a big evolution of technology, making elements smaller," said Hannu Laatikainen, an executive vice president of Murata, which sells accelerometers and gyroscopes to carmakers.

Don't expect too much skimping, though, because these sensors will be the foundation of life-and-death decisions. In cases where sensors come up with conflicting information, a computer could survey several sensors for a majority opinion. But the best situation would be to get accurate, reliable information at the outset.

This view from a Bosch self-driving car's laser-based lidar sensor shows distances to nearby objects such as cars to the left and right.
This view from a Bosch self-driving car's laser-based lidar sensor shows distances to nearby objects such as cars to the left and right. screenshot by Stephen Shankland/CNET

People accustomed to being in charge of a car might well be alarmed at the idea of ceding control to a computer. But it's also important to remember that computers have much better attention spans than humans. They don't get drowsy, and they can look in all directions at all times with multiple sensors. People will have time to get used to the idea, too, since the technological systems begin by augmenting human drivers rather than replacing them outright.

Bill Gross, CEO of startup incubator Idealab, said Google's self-driving cars slurp up 750 megabytes of data per second to do the job. That's a lot of data -- and equally important, a lot of data processing. Basing decisions on live data gathered in the moment lets the cars handle unexpected events such as pedestrians walking into the roadway and gradually take more responsibility.

Today, tilt sensors apply parking brakes automatically, stability control takes over aspects of steering and braking to keep cars from rolling over, and parking assistance puts parallel parking into the car's hands. Next comes traffic-jam assistance, where the car takes over accelerating, braking, and staying in the lane when in start-and-stop traffic. Later will come technology that helps cars change lanes or take exits. Eventually, the entire driving operation will become fully autonomous, Wallace predicted.

Sensors will spread throughout cars -- perhaps even to tires themselves. Today's cars have tire-pressure monitors built into valve stems, but they rely on batteries that can run out of power. Building them into tires could mean they could be self-powered through a process called energy harvesting, said Bert Gyselinckx, general manager of the Holst Centre R&D lab in the Netherlands.

"You put a harvester on one specific site on the surface inside a tire," and every time that part of the tire hits the road, a force about 100 times that of the Earth's gravity hits the harvester, Gyselinckx said. A capacitor, sensitive to the resulting pressures, generates a little electrical current that can be stored in a battery. It produces "a couple hundred microwatts" of power, enough to power a wireless transmitter to send information to the car control system.

Gyselinckx estimates that self-powered pressure monitors will be built into tires in about three years, with strain gauges arriving in six or seven.

Tire sensors can do more. Monitoring strains within each tire can show whether a car is accelerating, braking, or slipping, information that can be beamed to the car's electronic stability control system.

Bosch's lidar-based sensors give a 3D map of what's near a car and overlay that atop navigation data for the roads.
Bosch's lidar-based sensors give a 3D map of what's near a car and overlay that atop navigation data for the roads. screenshot by Stephen Shankland/CNET