Tesla Motors(s tsla) CEO Elon Musk thinks that autonomous vehicles that drive themselves may not be as far off in the future as the auto industry makes them out to be. The big sensor arrays that Google(s goog), Toyota(s tm) and Ford(s f) are testing in their autonomous car prototypes are a long ways from becoming commercially viable or affordable, Musk said in an interview with Bloomberg. But he said a scaled down version of such sensor systems relying on camera imaging could power a form of car autopilot that automates many of the ordinary tasks of driving.
“I like the word autopilot more than I like the word self-driving,” Musk said in the interview. “Self-driving sounds like it’s going to do something you don’t want it to do. Autopilot is a good thing to have in planes, and we should have it in cars.”
Musk told Bloomberg that he’s discussed the autopilot concept with Google’s engineers and even mentioned the possibility of Tesla and Google jointly developing such a system. But Musk later tweeted from his Twitter account that his comments to Bloomberg were meant to be off the cuff, and that Tesla and Google weren’t announcing any big new initiative.
“Creating an autopilot for cars at Tesla is an important, but not yet top priority,” Musk tweeted. “Still a few years from production.”
Musk’s notion of an inexpensive autonomous driving system is an interesting one as it does away with the complex light detection and ranging — or Lidar — sensors that use 360-degree scanning lasers to render a 3D construct of the world around the vehicle. You’ve seen such funky laser arrays on top of mapping vehicles before, so the technology is readily available, but even a scaled down version of such systems would add enormous cost to car. According to Ford engineers, those costs are among the principal reasons why fully autonomous cars are still many years away from becoming reality.
But Musk posited that a car could simply use cameras to collect the data the car needs from its surroundings. The approach has its limitations compared to lasers. Cameras take pictures, while Lidar is registering physical objects, tracking their distance and relative speed to your car. Cameras would extrapolate that data from images. Stereoscopic lenses could allow for better depth perception on the road, but ultimately a camera-based system would depend on a lot of image-processing muscle to cull critical life-saving information from a massive pool of pixels.
Luckily for Musk, he’s not the only one thinking this way. Many of the new advanced driver assistance systems (ADAS) systems that warn drivers of impending accidents or alert them to lane drifting are camera based. It’s also no coincidence that graphics card giant Nvidia(s nvda) is making a big play in the connected car space. The silicon vendor believes that its years of crunching real-time image and spatial data in PCs and game consoles make it the ideal company to power future car sensor systems.
Eventually cars won’t just sense the other vehicles and objects in their vicinity. Those vehicles will actively communicate with one another and those objects. Using a secure form of long-range Wi-Fi, a car will tell all the vehicles behind its about brake or change lanes. Transmitters on the highway will tell you a big curve is coming up. (For more info on the connected car see our infographic.)
Such vehicle-to-vehicle and vehicle-to-infrastructure communications could never replace sensors entirely since it would take decades for every vehicle on the road and every stretch of asphalt to become connected. But those technologies would allow cars to evolve cars beyond autonomy into a kind of ad hoc network, in which they would collectively make decisions instead of merely reacting to each other’s actions.
Returning to Musk’s remarks, though, I find it particularly interesting how he’s positioning the concept of the driverless car. He’s calling it a form of autopilot not too far off from the adaptive cruise control we use in our vehicles. Musk is downplaying the idea of car usurping control from the driver. He’s keyed in on the biggest obstacle facing the autonomous vehicle today: consumer perception. Few people are ready to cede act of driving to computer.