Accidents Show Smart Cars Need Wide Vision to Spot Cyclists
Automatic emergency braking systems need 180-degree frontal vision to help cars avoid hitting cyclists
Automatic emergency braking that can help cars avoid hitting pedestrians could become standard in many cars in the coming years. But a new study suggests such safety systems will need sensor coverage spanning almost 180 degrees in front of the car to avoid colliding with faster-moving cyclists.
Automatic emergency braking (AEB) systems have been offered as an optional upgrade for many cars since automaker Volvo introduced them in 2009. Many early AEB systems focused on avoiding collisions with cars and trucks, but more recent versions have expanded their safety capabilities to avoiding pedestrians.
Now U.K. researchers have turned to real-world accidents to look for lessons in how AEB systems could help cars avoid hitting cyclists who share the road.
“The study specifies the location of a representative sample of 175 pedestrians and 127 cyclists relative to the striking vehicle in the three seconds before impact,” said James Lenard, a research analyst at Datarye Ltd and visiting fellow at Loughborough University in the U.K. “This level of detail hasn’t to my knowledge previously been released into the public domain.”
Lenard and his colleagues found that “almost complete 180-degree forward vision is required to provide a 90 percent detection zone for cyclists.” By comparison, collisions with pedestrians rarely occurred outside a frontal viewing cone of just 35 degrees. Details of the study were published on 19 March 2018 in the journal
Accident Analysis & Prevention
Much of the difference in the potential angles for collision with pedestrians and cyclists is based on the finding that fast-moving cyclists cover more ground in the three seconds prior to impact. The combination of cyclist speeds and similarly fast-moving cars meant that cyclists could be scattered much more widely in front of a car’s field of view before accidents occurred.
Such knowledge could prove crucial as safety experts at the European New Car Assessment Programme (Euro NCAP) are developing safety rating tests for AEB systems. Across the Atlantic, 10 top automakers have also pledged to
make automatic emergency braking systems standard
in all future car models sold in the United States.
The rise of automatic emergency braking, also known as autonomous emergency braking, offers an opportunity to make manual driving much safer long before the advent of self-driving cars. But current examples of such AEB systems have different capabilities to detect and respond to vehicles, pedestrians, and cyclists.
“Based on my knowledge of information in the public domain, it would appear that production AEB systems will evolve towards 180-degree frontal coverage in the future, especially if consumer ratings systems such as EuroNCAP provide an incentive,” Lenard said.
Knowing the differences in potential car collisions with pedestrians and cyclists could help engineers design more efficient AEB systems and other safety features in future self-driving cars. “Beyond 35 degrees, the priority is to determine whether any figure provisionally identified as a potential human being is a cyclist, since impacts with pedestrians out of this region are rare,” Lenard pointed out.
The U.K. accident data also showed that cars tend to encounter many more cyclists than pedestrians at road intersections. That might lead engineers to design a car’s software to prioritize keeping a lookout for cyclists coming from the side when a car is edging into an intersection or roundabout.
The results could even prove helpful for self-driving cars that usually have sensors covering a 360-degree field of view. That’s because such research could “help prioritize the operation of the image-recognition and decision-making software, assuming that not every reading flowing through the system from every sensor can be fully processed in real time,” Lenard said.
Lenard and his colleagues drew upon a unique accident dataset that was originally compiled for the U.K. Department for Transport. Their study’s focus on extracting time-to-collision data received funding from the U.K.’s Thatcham Research and the U.S. Insurance Institute for Highway Safety.
Such U.K. accident data could provide an important baseline for other countries. Still, Lenard acknowledged there could be variations in accident conditions across different countries and even between urban and rural areas. Regions very different from the U.K. would likely need their own studies involving detailed accident data.
“In countries with more dedicated bicycle lanes, it is conceivable that cars turning across the path of ‘undertaking’ bicycles might be more prevalent than in Britain: this could imply more AEB ‘rear vision,’” Lenard explained. “In countries with higher impact speeds than Britain, a longer [detection] range might be beneficial.”
The world does not need to wait for companies to perfect self-driving cars in order to improve road safety. Different versions of automatic emergency braking systems already exist and could become more commonplace through either regulatory standards or market demand.
Some experimental self-driving cars may not even meet the safety standards offered by commercial AEB systems, as seen in the recent death of a pedestrian in a self-driving Uber accident. A “contemporary production pedestrian AEB system” should have detected the pedestrian walking her bicycle across the road, Lenard said.
In any case, Lenard suggested that the U.K. study can help development of advanced car technologies beyond just automatic emergency braking. If such braking systems require “lateral vision” to spot cyclists, then the same logic would apply for technologies such as lane-change assist and side-impact pre-activation.
“The study illustrates that only a holistic approach to the advanced technologies makes any sense,” Lenard said. “
Manufacturers and suppliers will consider whether an integrated set of sensors and software provides the most efficient and cost-effective solution for all of these functions, rather than work through them independently one by one.”