Many new cars today are equipped with advanced driver assistance features. They help with driving or parking and include features such as automatic emergency braking and forward-collision warning. Sometimes cars are marketed as self-driving, but most actually just have driver assistance technology.

Since companies like Tesla haven’t completely ironed out the glitches in their self-driving cars yet, it could take several years for autonomous cars to be safe enough to enter the market. But which is safer? We look at the safety mechanisms and safety improvements for both as well as which one offers the most safety potential.

Safety mechanisms differ depending on the degree of automation

According to the National Highway Traffic Safety Administration, six levels of driver assistance technology have been established by the Society of Automotive Engineering (SAE). They range from Level 0 with no automation to Level 5 that consists of full automation. SAE has defined the extent of human involvement for each level.

Most production vehicles built today fall somewhere between Level 0 and Level 2, which covers driver assistance and partial automation. Cars produced in the past five years are more likely to have Level 1 tech, which helps drivers with steering and braking/accelerating but not both at the same time. Level 2 or partial automation tech combines steering and accelerating functions at the same time and requires driver monitoring. Levels 1 and 2 include cars with driver assistance features.

A self-driving car such as Google’s Waymo operates at Level 4, high automation. The car can perform all driving tasks under certain conditions but allow the driver to intervene when necessary. The goal for Google, Tesla, Ford, Uber, and other automakers that are working on autonomous cars is to reach Level 5, but no car company has reached this goal yet.

The safety mechanism for cars with driver assistance tech or partial automation relates to what SAE calls functional safety. The goal is for the car’s safety equipment to function properly while reducing the risk of injury or death to the driver and damage to property. This safety mechanism is meant to focus on preventing failure in the electric and electronic components in the system.

Self-driving cars have a safety mechanism that is based on the “safety of the intended functionality” or SOTIF. It applies to artificial intelligence and machine learning used to operate autonomous cars. This safety mechanism requires that developers create automotive software that accounts for trains the car’s computer to respond to a wide range of hazardous situations in order to keep drivers safe. 

Of course, engineers who are designing a self-driving car will build the SOTIF safety mechanism on top of the functional safety mechanism. But the two safety mechanisms are distinctly different. Functional safety is risk-based and is based on the avoidance of system equipment failure, while SOTIF is software-driven and proactive in training the onboard computer to sense real-time driving dangers.

Safety improvements between self-driving cars and driver-assist tech also differ 

The measure for safety improvements for cars equipped with advanced driver assistance systems is how much these systems can reduce crashes. The National Safety Council states that 94 percent of serious crashes are caused by driver error. It cites 2019 data from the Insurance Institute for Highway Safety that shows forward-collision prevention features reducing front-to-rear crashes with injuries by 56 percent.

Lane-keeping assistance systems reduced single-vehicle, sideswipe, and head-on crashes with injuries by 21 percent. Blind-spot detection reduced crashes with injuries by 23 percent, and backing collision avoidance systems reduced crashes with injuries by 78 percent. No data for pedestrian detection systems is available yet.

In comparison, safety improvements for self-driving cars are based on the data gathered by the miles the car has traveled. The further self-driving cars travel, the more the car’s computer learns about its driving environment. As a result, safety engineers look at the distance in miles between human interventions, crashes, and fatalities to measure safety improvements for these cars.

It’s hard to get a sense of how safety has improved in these cars are because relatively few of them travel on the road and none have Level 5 autonomy. Also, they will ultimately require millions of miles of driving to quantify crash reduction, according to a report by the Rand Corporation.

The highly-publicized crashes that involved Teslas using the AutoPilot tech and the accounts of the deaths caused by crashes in similar vehicles muddy the issue of safety improvement for autonomous or even semi-autonomous cars even more. 

Does self-driving technology have the most safety potential?

Both systems require that drivers use good judgment and stay alert and engaged in order to stay safe. People who drive cars with driver assistance systems can’t let their guard down and assume that the systems will protect them in every situation. And people driving autonomous cars will have to seek a balance between allowing the car to do what it’s programmed to do and intervening when necessary.

But if we look at the big picture, autonomous cars have enormous safety potential. They have the potential to keep distracted, tired, or drunk drivers safe. They may also offer new mobility and independence for elderly and disabled people. These cars may perform better than human drivers due to improved perception, decision-making, and vehicle control. 

Self-driving cars can compensate greatly for human error and may be part of the solution to reducing the 37,000 deaths caused annually by automobile accidents. But there are two questions we should think about when looking at their safety potential. One is: how safe is safe? If engineers can train a car’s computer to keep its driver safe, have they been able to come up with every driving situation that presents danger?

The other question that comes up about autonomous cars is, how many years will it take to put a safe Level 5 car on the road? Validation and testing for these cars could take years. 

Although autonomous cars probably will offer more safety potential than those with driver assistance systems in the future, we still need cars with driver assistance systems to keep drivers safe right now. The technology is in place in these cars to reduce crashes and related injuries as well as prevent deaths, and it is continuing to improve each year.