Driverless Cars Still Can’t Beat Mother Nature

Posted by

As modern driver aids begin to watch the road ahead, they run into challenges from almost every weather condition.

Every automaker is quickly moving to make automatic emergency braking, adaptive cruise control, and lane keep assist standard on all their top-selling models. Toyota leads the way with most of its vehicles having the technology standard on almost all models and all trims in 2018.

The industry has committed to making the technology standard on all of the vehicles produced in a few years. We love the technology and it has saved our bacon in real-world use.  However the more we test these systems, the more we discover their limitations. These systems all form the backbone of self-driving vehicle technology and without solutions, autonomous vehicles will never work right in many conditions.

Looking for a new or used vehicle? Start your search at

One of the most common situations in which new automatic emergency braking systems are disabled is slush. When driving near the freezing point of rain and snow, the front of the vehicle becomes caked with the frozen gloop thrown up from the cars ahead. That ice-snow-dirt frizzle coats the radar sensor mounted on the vehicle’s grill logo and it then stops working.

It is easy to understand why a system might be disabled by ice and snow building up on the sensor. But some of the highest-ranked automatic emergency braking systems can be disabled by a bright sunny day. Subaru has one of the industry’s top-rated optical safety systems. Using two cameras, the EyeSight system looks ahead and can stop the vehicle without driver involvement if it detects an obstacle ahead. However, anyone with the system who has a commute that heads directly east in the morning or west in the afternoon knows that the system can’t see in direct sunlight, shuts off, and displays a warning sign that the driver is now own their own. Just like the old days.

Driver aids like adaptive cruise control can have trouble where they seem to make the most sense. For example, on steep hills. Anyone who has driven in thick fog knows that it is one of the scariest situations of all. A system that can brake better than a human operator and see the road ahead more clearly would be very handy. Yet, these systems don’t operate well in fog.



Other natural situations that can disable driver aids and safety systems include smoke and dust. Chips and cracks on a windshield from road salt and sand up high where the optical sensors reside can also cause them to stop working properly.

The current solution to autonomous systems thwarted by weather is LiDaR, an acronym for “Light Detection and Ranging,” as opposed to RaDaR, an acronym for “Radio Detection and Ranging.”

Lidar (we’ll dispense with the fussy capitalization from here on out) has several distinct advantages over radar. Lidar sends out millions of laser pulses per second to build a real-time map of the road ahead. Lidar on its own isn’t smart enough to recognize snowflakes and raindrops as anything other than obstructions, but matched with algorithms like those developed by Ford and the University of Michgan, the system can ignore the weather and concentrate on the road and traffic hazards ahead.

Which would be great, if lidar was suitable for anything but experimental technology in its current form. For fleets like Waymo’s it works because the lidar system is built on top of a Pacifica minivan with plenty of roof area. That Waymo fleet is also an experimental idea backed by venture capitalists with money to burn. Nobody really cares about whether it’s going to make money or be profitable in the early going. It’s more important to be first.

But large, unaerodynamic lidar arrays aren’t suitable at all for a fleet of vehicles sold by a major auto manufacturer.

First of all, they’re still wildly expensive. The Velodyne HDL-64E lidar sensor on Google’s car, for example, cost $75,000. The sensors on Waymo’s fleet of Pacificas are the Velodyne VLP-16 or “PUCK,” and Google — via Waymo — has singlehandedly invested enough in the technology to drive the price down to about $8,000.

But $8,000 added to the COST — not the MSRP — of a new car is enough to send every accountant in the automotive industry straight to rehab. It’s simply not feasible at the moment to consider lidar as a viable option for the general public until the cost of that technology is much more in line with that of radar.

Second, manufacturers are working day and night to make vehicles as slippery and light as possible to increase fleet fuel economy. Sticking a heavy lidar array on the roof, protruding into the wind, is at cross-purposes with that goal.

For the near term, we’re stuck with line-of-sight technologies that have existed on cars for at least five years now. With every passing year that another inexpensive, reliable technology that can see through the weather doesn’t materialize, we’re further down the road to Level 4 Autonomy in your average daily driver.