According to the National Highway Traffic Safety Administration (NHTSA), 9 out of 10 crashes tend to be the result of human error. Driver mistakes are a major factor in the failure chain of events leading up to a vehicle accident. The Insurance Institute for Highway Safety (IIHS) recently published a study that evaluated how many crashes would likely be reduced if automated vehicles were to replace a significant portion of human drivers on public roadways. The result from the study was that only about one-third of accidents involving light duty vehicles would be reduced, while the other two-thirds were dependent on the level of safety of the automated driving system (ADS).
Specifically, the study examined 5,000 police reports collected from NHTSA. The crashes examined had two primary requirements: there must have been a vehicle towed and Emergency Medical Services (EMS) must have been called to the scene. These requirements are an important distinction in determining the severity of an accident; those with EMS response and that require towing are likely to be more significant than other crashes. IIHS was able to categorize the 5,000 police reports into 5 primary categories related to the factors leading up to the crash.
Categories of Crashes
- Sensing/Perception: This relates to how well the human was able to identify relevant objects and other road users. The crashes in this category involve scenarios where the human was unable to appropriately identify the object. For example, it could be raining heavily and the human could not see another road user in front of them.
- Prediction: This specifies how well the human could predict the behavior of another road user or object. An example of this could be scenarios where the human was able to identify another road user, but the human made an inaccurate prediction as to the direction or velocity of the object. This happens frequently with individuals trying to cross the street and drivers inaccurately understanding their relative velocity and intended heading.
- Planning/Deciding: This involves poor decision making on the human. The driver may have been able to identify a road user and predict their motion, but then made a poor decision based on that information. For example, the human may have decided to speed past a yellow light even if they noticed someone trying to cross the street at the same time.
- Execution/Performance: This relates to scenarios where the human did not perform the correct action in the driver’s seat. For example, maybe the wheel slipped from their grip or they stepped on the braking pedal too little.
- Incapacitation: This final category involves accidents where the human was either asleep, drunk, or unable to appropriately execute the dynamic driving task (DDT) in some form or manner.
Significance of Results
Altogether, the study found that the incapacitation cases would absolutely be solved by introducing automated vehicles. However, for the other categories to be reduced, automated driving systems must prioritize safety over convenience or time management. The good news is that Original Equipment Manufacturers (OEMs) continue to integrate safety as part of their development and testing on a regular basis. The Automated Vehicle Safety Consortium is continuing to work with manufacturers to develop best practices related to safety. There is a ton of work happening in this space, and QS-2 is continuing to provide advanced hazard analysis and functional safety expertise for different groups working on automated vehicles. We look forward to building public trust in this new technology and helping share how safety is being integrated into everything OEMs are doing!