Image: Just Super/Shutterstock
There’s nothing magic about it, but a new technology could massively improve the safety of driverless cars.
Before driverless cars can consider taking to the road commercially, developers are trying to figure out those little things we take for granted as human drivers that make us adept at driving.
But, in some instances, technology currently being developed even outshines what we are able to do and touches on something that almost sounds magic in nature, but is really just advanced science.
One such breakthrough was achieved recently by researchers from Stanford University who have found a way to let a driverless car ‘see’ around a corner to predict when to make an immediate stop – for example, if a child runs out from behind a winding street.
The research team described how it set a laser next to a highly sensitive photon detector capable of detecting the smallest detail, down to a particle of light.
Then, on the other side of a physical partition, the researchers shot particles of light from a laser against a wall that would bounce off and be picked up by the detector on the other side.
At its current level of capability, this scan can take between two minutes and an hour, depending on conditions such as lighting and the reflectivity of the hidden object.
When the initial scan finishes, an algorithm untangles the paths of the captured photons and combines them again to create a much sharper image of what’s behind the partition.
It is believed that the process could be speeded up to become almost instantaneous once the scan is complete.
Better LiDAR
Some of the complications holding it back so far include the fact that the distance to the object and amount of ambient light can make it difficult for the team’s technology to see the light particles it needs to resolve out-of-sight objects.
Current object detection systems in driverless cars, such as LiDAR, intentionally ignore scattered light particles in order to create a clearer picture of what’s in front of it, but that might not be enough.
“We believe the computation algorithm is already ready for LiDAR systems,” said Matthew O’Toole, co-lead author of the paper. “The key question is if the current hardware of LiDAR systems supports this type of imaging.”
The system also needs to be able to work better in daylight with moving objects as, until now, tests had only been conducted with indirect light.
The team is very hopeful, however, that these obstacles can soon be overcome.
