Autonomous vehicle braking: still a bit hit and miss

Auto-braking system

Many new cars automatically brake before striking another vehicle in their path. Though automakers have coined different names for this new feature, all of these systems operate in much the same way: they use a combination of cameras, radar and lidar to “look” ahead for obstacles and then apply the brakes to avoid or lessen an impact when the driver doesn’t respond in time.

For our first tests, we checked if the size of the radar signature and the contrast of the vehicle outline affected the system’s performance (cite Yang et al). They did not, which means that the system is robust within the range of conditions we tested. During our tests, we also saw that driver interaction with the gas and brake pedals during the AEB response sometimes interrupted the system and resulted in an impact that would not have happened without the interruption. These findings highlight the difficulties these systems confront when deciding if and how to return vehicle control to the driver once the system has engaged.

In another set of tests, we saw that the system worked well when accelerating from rest towards a stopped car provided a speed of 25 km/h was reached (cite Xing et al). At lower speeds—including the IIHS test speed of 20 km/h—the AEB system did not engage consistently and the stopped car was hit. Since the Toyota AEB worked flawlessly in the IIHS tests, our finding highlights a potential shortcoming of the constant speed run-up used in the IIHS tests. Consumers might expect these systems to work at low speeds in stop-and-go traffic, but our tests indicate they may not.

Overall, we found that Toyota’s AEB system performed as expected most of the time. In some special circumstances, the system did not prevent an impact, although in most of these cases, the severity of the impact was lower than if the system was not present. Many more variables and many different systems still need testing, but these initial tests tell us that autonomous vehicle braking is a complex task and that some real-world situations may foil these systems.

related expertise

related professionals