People are, on average, terrible drivers. How do you solve this problem? Do not let them drive or put a machine in its place?
Self-driving vehicles are not some futuristic automotive technology that we are yet to see. These are cars from the likes of BMW and Tesla with self-driving features on the road. There are two types that are being tested or that are on the road: semi-autonomous cars that we already see and fully autonomous ones that are expected anywhere between two to 14 years depending on whom you ask. Tesla CEO Elon Musk and Google’s Sergey Brin are two people who believe that it will be sooner rather than later.
Chris Urmson, who heads Google’s driverless car programme, said in a talk, “My oldest son is 11, and that means in four and a half years he will be eligible to get a driver’s licence. My team and I are committed to make sure that does not happen.”
So when these arrive, what would enable cars to safely roam our streets by themselves?
Sense of sight
In order for your car to be aware of its surroundings, it first needs to see where it is standing.
LiDAR. LiDAR, or Light Detection And Ranging, is one of the most popular and actively-used technologies that enable self-driving cars to see. The most recognised version of LiDAR is the one you see on top of Google’s self-driving cars, and it is HDL-64E.
Velodyne’s HDL-64E enables obstacle detection and navigation with a 360° field-of-view.
Enabled by 64 laser beams, this is probably one of the best and most expensive solutions available right now.
Another firm, Quanergy, recently announced a solid-state LiDAR named S3 that has no moving parts or micro-mirrors. Instead, it uses an optical-phased array as a transmitter, which can steer pulses of light by shifting the phase of a laser pulse as it is projected through the array. S3 is due to be available in September.
The major benefit of using LiDAR systems is that the data these generate is easier to compute and process as compared to the data generated from a camera. These can also work in places where GPS and mapping data are not really accurate. Volvo Group has an autonomous FMX heavy-duty vehicle designed to work in underground mines without the need for a driver.
Of all the demonstrations that we at EFY have come across of LiDAR usage, perhaps the best is when the car picks up a cyclist waiting to cross the road on the other side of a pickup truck. Drivers in the car are not able to see the cyclist, but LiDAR laser picks it up, calculates the distance and prevents the car from moving ahead until the cyclist has passed.
Cameras. Mercedes F015, about which we wrote last year, does not use LiDAR, but a mix of 3D stereoscopic cameras, radars and ultrasonic sensors for its autonomous functions. It is very much possible to set up driver-assistance systems using just camera data, like how Land Rover Discovery Sport does with a Bosch stereo video camera. Its spatial measurements and video signals provide enough data to calculate, for example, the distance to vehicles ahead.
This is nothing new, of course, as Subaru has been marketing their own version since 2010, and Continental has their own forward-looking braking system with stereoscopic cameras that is just as old.
Infrared or thermal imaging cameras have been available that can detect hidden objects like people or animals in total darkness, smoke or fog-like conditions when normal detection systems may not easily pick these up. FLIR Systems PathFindIR is a compact thermal imaging camera that significantly reduces the hazards of night time driving, and is designed for both military vehicles as well as commercial vehicles as an aftermarket part. These systems have improved from mere night vision to being able to identify a hazardous object or even go as far as alerting the object.
Mercedes had CL550 a few years back that would identify pedestrians on the road and flash them with four blasts of diverted headlight while highlighting them with white brackets on users’ screens.
While LiDAR and cameras are great for generating an accurate map of the car’s surroundings, these are not ideal for monitoring the speed of other moving objects in real-time. So what do we do?
