Self-driving cars on the roads is increasing, but how does a self-driving car work – how does it know when to stop, or if someone is about to cross the road?
Driverless vehicles are equipped with a range of sensors to help them navigate the roads and are critical for detecting dangers and simply safely driving on the roads. They also are responsible for prompting an appropriate action, such as an evasive maneuver, decelerating to turn a corner or stopping.
The sensors fall into three categories – cameras, RADAR and LiDAR, each of which is explored below.
Cameras are among the cheapest of sensor technologies and consist of series of CMOS (complementary metal-oxide semiconductors, basically a chip that stores information) imaging sensors to produce images of between 1 and 2 megapixels.
Rear and 360° cameras positioned on the front and back of the vehicle give drivers a representation of the environment outside the vehicle while also offering a wide range of functions from evaluating speed and distance to determining objects from their outlines.
Most cameras are 2D but some manufacturers – especially luxury class car makers – are beginning to incorporate 3D cameras too. Both 2D and 3D camera need sensors with a very high dynamic range of more than 130 dB, which is necessary to ensure a clear image even with direct sunlight on the lens.
Radar or radio detection and ranging – uses radio waves to detect and localize objects. Electromagnetic waves are emitted, hit and bounce back from an object, and are detected upon their return, to provide information about an object’s location and speed.
Current systems are based on short-wave sensors at 24 GHz or longer wave at 77 GHz. Both sensors are used at the front and back of the vehicle to monitor surrounding traffic ranging from a just few centimeters to hundreds of meters away. The shorter-range sensors are found on the corner of the vehicle and are designed to support highly automated driving; as such they are generally used for blind spot detections, parking assist and emergency braking for example. The longer-range sensors are forward facing and placed at the front of the vehicle; they offer higher accuracy for speed and distance measurements, more precise angular resolution, have a smaller antenna size and therefore lower interference problems.
LiDAR - or LIght Detection And Ranging – was developed shortly after the advent of lasers in the 1960s, and was first used to measure clouds. It is perhaps the most significant piece of hardware in the contest to develop self-driving cars. LiDAR shoots rapid pulses of ultraviolet (UV), visible or near infrared (IR) laser light at a surface – up to 150,000 pulses a second – and measures the time taken for the pulse to bounce back from the target. The distance to the object is deduced by using the speed of light to accurately calculate the distance travelled. The result is an accurate three-dimensional picture of the target object and its surface characteristics.
Google and Uber are among the many names developing autonomous vehicles: their roofs feature a continuously spinning box which gives 360° visibility and precise, in-depth data about the exact distance to an object to an accuracy of ±2 cm. This box is the LIDAR system and consists of a laser, scanner and optics and a specialized GPS receiver, which is especially important if the system is moving. Huge 3D maps are generated for the vehicle to navigate through, while pedestrians and cyclists, traffic signs and other nearby obstacles are also detected.
However each sensor technology is not without its pitfalls and its likely a combination of the three will be necessary to make a safe, truly autonomous vehicle.
References and Further Reading