The basic building blocks behind all autonomous and ADAS systems are sensors which act as the vehicle’s “senses”. They enable the development of perception and provide information on the vehicle’s environment. Such sensors include radars, cameras, lidars, inertial sensors, and GNSS signal receivers. The properties of each sensor help understand and navigate the vehicle’s environment better.
Sensors can be divided into passive and active. The former does not interfere with the external world and are generally more resistant to harsh conditions. These include inertial sensors (IMU), for example. They measure inertial forces influencing the system, i.e., accelerations and angular velocities. They are often equipped with a magnetometer that measures the magnetic field strength, which can be used to estimate the vehicle orientation based on the Earth’s magnetic field. It should be noted, however, that these measurements are not always reliable and often require appropriate calibration. Another passive sensor is a camera, which receives electromagnetic waves in the visible or infrared band and streams the image captured by the device’s image sensor. A single camera provides a flat representation of its field of view. A combination of two cameras, however, provides a stereo image that also includes information on image depth. Another important passive sensor is GNSS, i.e., satellite receivers that help determine the location in a global coordinate system. The most popular GNSS system is GPS, but there are other systems as well, including GLONASS and Galileo. Although there are techniques to increase the accuracy of positions determined by these sensors, such as RTK or DGPS, atmospheric interference, dense buildings, tunnels, and halls are still problematic as they significantly reduce signal quality and thus positioning accuracy. Passive receivers also include odometric sensors, which can be used to determine wheel rotation speed, as well as pressure and humidity sensors, which can be helpful in correcting the readings from other sensors. Active sensors emit an additional signal to the environment, and then receive it to measure a given value. Popular active sensors include ultrasonic sensors, which emit an ultrasonic sound pulse and measure the time needed to receive it. Thanks to their simplicity and low cost, they are often used in parking assistance systems. Due to their short range and low accuracy, they are not suitable for faithful and more precise gathering of information on the vehicle’s environment. Others popular active sensors include radars, which emit a radar beam that returns to the sensor receiver after bouncing off an obstacle. By measuring the time between the sending and receiving of the beam, the distance from the obstacle can be determined. Meanwhile, the change in frequency of the transmitted beam is used to determine the radial speed based on the Doppler effect. Even though this mechanism is successfully used in Advanced Driver Assistance Systems (ADAS), the radar signal is often too inaccurate to provide a reliable image of the environment, unfortunately. Not every obstacle reflects the radar beam in the same way, so pedestrian detection may not provide reasonable results. Another problem is the reflection of signals in built-up areas, including tunnels or heavy traffic. A next-generation active sensor that is revolutionizing the autonomous vehicle market is LiDAR (Light Detection and Ranging). Its principle of operation is similar to radar, but it emits a laser beam in the infrared band. ToF (Time of Flight) lidars are highly precise, and they perform measurements just like a radar, i.e., measuring the time between a laser beam is emitted and captured by the sensor receiver. Unlike a radar beam, a laser beam is focused, thus has a higher resolution and enables a more accurate 3D mapping of the environment in the sensor’s field of view in the form of a so-called point cloud.
There are also active options when it comes to cameras, i.e., ToF and 3D cameras. They are simpler and cheaper alternatives to lidars, whereby the camera is equipped with an emitter projecting a point grid in infrared and a receiver which adds spatial information to the flat image based on the time it takes for the beam to return or its geometry.