Jeff Bier’s Impulse Response—Quietly, Our Cars Are Learning to Drive Themselves

Submitted by Jeff Bier on Mon, 08/25/2014 - 22:01

Lately there's been a great deal of buzz about self-driving cars. The idea of autonomous vehicles evokes excitement in some people and terror in others. Clearly, there are significant challenges to be overcome before self-driving cars can become mainstream.

To me, the most exciting aspect of self-driving cars is that many of the same technologies required for future self-driving cars are making human-driven cars safer today. Take Mobileye, for example. The Israeli company has been a pioneer in what the industry calls "Advanced Driver Assistance Systems" ("ADAS" for short). These systems use computer vision to reduce forward collisions, warn drivers when they are drifting out of their lanes, and read road signs to keep drivers appraised of speed limits. Not surprisingly, such systems (from Mobileye and other suppliers) were first deployed by luxury brands like BWM and Cadillac. But in the past year or two, these features have started to become commonplace in mid-market brands like Honda, Ford and Subaru. Mobileye's remarkably successful initial public stock offering earlier this month is a testament to both the value and the growing adoption of its technology.

I recently drove a 2013 Cadillac XTS, and was impressed by the reliability and ease of use of its ADAS capabilities. This car is equipped with a number of ADAS functions, including lane departure warning. If the car begins to drift out of its lane, the driver is alerted with a vibration on the left or right side of the seat, depending on the direction of drift. Current models from Subaru, Acura and Mercedes take this idea a step further, automatically steering the car to keep it centered in its lane. Combine this lane-keeping capability with adaptive cruise control (when adjusts the car's speed based on the speed of the vehicle ahead) and automatic braking, and you have a car that actually can drive itself under certain conditions, as this video demonstrates.

Over three years ago, I had an early Mobileye system retrofitted into a car in order to gain some first-hand experience with computer vision-based ADAS. This system has been useful, but its usability is hindered by the fact that it is not fully integrated with the vehicle's other systems. It also suffers from too many false alarms. For example, with the system I installed in 2011, once or twice a day, when driving on suburban streets, a very loud, shrill alarm will erroneously warn of an impending collision. This is more than an annoyance, because it means that the driver must learn not to react too quickly to the system's warnings, instead taking time to assess the situation. This significantly detracts from the system's effectiveness for non-highway driving and tempts the driver to switch it off. (On highways, presumably due to their more uniform structure, I haven't experienced false alarms.)

In contrast, the Cadillac system, which is fully integrated into the vehicle, was a pleasure to use, and was remarkably reliable. In a week of driving the Cadillac several hundred miles on highways, suburban streets and country roads, I did not experience a single false alarm. The improved usability and reliability of the Cadillac's driver assist system has made me a believer: Regardless of how long it takes for ultra-safe self-driving cars to become commonplace, well-designed driver-assist systems can make human-driven cars safer today. That's a very exciting prospect, considering that over one million people die annually in automobile accidents worldwide. It's also a big market, considering that the annual production of motor vehicles is approaching 100 million units per year.

Have you driven a car with vision-based ADAS? If so, I'd love to hear about your experience. Post a comment here or send me your feedback at http://www.BDTI.com/Contact.

Jeff Bier is president of BDTI and founder of the Embedded Vision Alliance.

Add new comment

Log in or register to post comments