Study Finds Autonomous Cars Cannot Prevent Most Crashes
A new study released by the Insurance Institute for Highway Safety (IIHS) states that the current capabilities of autonomous cars will prevent only about a third of all collisions. While self-driving car technology would be able to correct for some human error, the study finds, it would not be able to address the bulk of situations that cause collisions.
One of the big promises of self-driving cars is the prospect of safe travel, with no fear of human error, resulting in fatality-free commutes and road trips. Drivers, whose mistakes or failure in judgement account for 9 out of 10 collisions, would relegate their role of making thousands of decisions every minute, and entrust their lives to computers that don’t get tired, distracted, or drunk. It is a lofty goal into which car companies are investing billions of dollars.
This new study, however, suggests that it could take billions more to achieve success.
The study analyzed 5,000 crash reports and found five driver-related factors that can cause collisions. They are, as stated by the study:
- “Sensing and perceiving” errors. These include things like driver distraction, blocked visibility, and failing to recognize hazards before it is too late to respond.
- “Predicting” errors. These occur when drivers misjudge a gap in traffic, incorrectly estimate how fast another vehicle is going, or make an incorrect assumption about what another road user is going to do.
- “Planning and deciding” errors. These include driving too fast or too slow for the road conditions, driving aggressively, or leaving too little following distance from the vehicle ahead.
- “Execution and performance” errors. These include inadequate or incorrect evasive maneuvers, driver overcompensation, and other mistakes in controlling the vehicle.
- “Incapacitation,” which includes impairment due to alcohol or drug use, medical problems, or falling asleep at the wheel.
Based on the study, the IIHS found that crashes involving “sensing and perceiving” errors and “incapacitation” comprised about 34% of accidents, as long as all vehicles systems were in perfect working order. These are the types of crashes that autonomous driving technology can avoid.
The study also found that “predicting,” “planning and deciding,” and “execution and performance” errors contribute to 40% of collisions. According to the IIHS, these types of crashes might still occur because self-driving technology does not yet exist that addresses these issues.
For example, as a driver, you might be able to observe a stationary pedestrian and “sense” whether the pedestrian will attempt to cross in front of your vehicle. Right now, driving assistance technology doesn’t have that same capability.
Here is another example. Mechanical failures, such as a tire blowout, can still cause a collision whether a human driver is in control or the software of a self-driving technology.
As of now, existing semi-autonomous driving systems are designed to place a driver’s decisions above those of the computer. With adaptive cruise control, for instance, a driver may choose to cruise at a speed that is above the posted limit, overriding the car’s programming to stay at or under the limit. Or, as with lane keeping assistance, a driver can override the automated steering that nudges the car back into the lane.
In order to reach a higher safety record, concludes the IIHS, the car’s software will need to be able to override a driver’s preferences.
Clearly, a fully autonomous and crash-free future is still in the building stages.
The information in this article is from the Insurance Institute of Highway Safety. It was accurate on June 5, 2020 but may have changed since that date.