Sensors for Autonomous Vehicles

We have discussed several sensors that can be added to image sensors to make our autonomous vehicles safer and more reliable. We can look at some of the sensors, their characteristics and how they can be utilised in an autonomous vehicle.

When it comes to sensor selection, one will have to choose from key technical characteristics like the sensor’s:

1. Sensitivity, which is the minimum value that the sensor can detect.
2. Resolution, which is the minimum variance between any two measurements.
3. Accuracy, which is the error between the true value and the sensor’s reported measurement.
4. Perspective, which is conveyed as the field of view.
5. Dynamic Range, which is the minimum and maximum values that can be accurately reported.
6. Timescale, which is the refresh rate or the frequency of the sensor. The sensor may also be active or passive. An active sensor emits a form of energy to sense the environment, while a passive sensor relies on ambient conditions to provide information.

Modern automobiles which aren’t autonomous already have some useful systems and sensors that can be utilised or increased in autonomous vehicles. The wheel speed sensor located at each wheel and outputs the speed of the wheel is used by existing Anti-lock Braking System (ABS) and cruise control systems in modern automobiles and will be needed in autonomous vehicles as well.

Other existing sensors include the Yaw Rate sensor, a gyroscopic device necessary for electronic stability control, and the Lateral/Longitudinal sensors, used for collision detection as well as electronic stability control.

Global Position System (GPS), which utilises a receiver and an antenna to communicate with various satellites for trilateration (very similar to triangulation) to establish the vehicle’s absolute position, is also available in most modern vehicles but not with the accuracy required for an autonomous vehicle. In autonomous vehicles, the latitude-longitude-altitude data from the GPS can be used to compute optimal routes, driving directions and even lane mapping.

GPS technology is based on Radio Frequency (RF) signals propagated from a satellite to a GPS antenna to approximate the position anywhere in the sphere of the signal propagation. Thus, with signals from a single satellite, an autonomous vehicle’s location on a sphere with the radius r (same as the distance to the satellite) can be established. If there are signals from two satellites, then the location problem is reduced into a circle. However, with signals from three satellites, it is possible to compute the absolute location from two known points. This assumes that the distances between the satellites are known.

The approximate distance from a single satellite and three satellites.

Sensors Required in Autonomous Vehicles

The following are external sensors that may not be present in modern automobiles but required in autonomous vehicles:

• Radio Detection and Ranging (Radar) can be used in autonomous vehicles for the detection of both far and near obstacles. Radar systems function very well in a wide range of environmental conditions and are generally immune to high luminosity, rain, fog, snow and even dust. A radar system propagates RF waves in the 76-77GHz range band and uses a receiving antenna to pick up returning waves, which are then amplified and filtered for noise. Adaptive cruise control systems typically utilise radar because they can detect distant objects travelling at high speeds.
• Light Detection and Ranging (LiDAR), compared to radar, uses a beam of light typically with an infrared laser diode that is reflected off a rotating mirror. When the light hits a non-absorbing object, it reflects to the sensor and creates a map of the object. High-end LiDAR systems make use of multiple beams of light to provide multiple distance measures simultaneously for a 3D model of the environment. LiDAR systems are typically not as accurate as RADAR systems in detecting speed of moving objects.
• RGB cameras are one of the heavily used sensors in autonomous vehicles. They are used for reading road signs and for lane departure algorithms, which may use the dynamic range of the camera to create an optimal contrast between the black in the road and the white or yellow lane markings. Camera systems generally absorb light that bounces off objects and the beam of light is separated into the three colour channels (red, green, and blue) before being fed into a complementary metal oxide semiconductor or a charge coupled sensor.
• Sound Navigation and Ranging (SONAR), reflects sound waves from various objects in range, and the frequency of the returned pulses is used as indication of the object’s distance. Sound waves require a medium to propagate and unlike an electromagnetic wave, sound is heavily affected by the medium. Hence as environmental conditions like temperature and humidity varies, so does the sensing capabilities of an ultrasonic sensor. Ultrasonic systems tend to be the cheapest compared to LiDAR, RADAR and Camera, and are generally used for parking sensors and other close-range applications in automobiles.

All these sensors are needed on a larger scale to make an autonomous vehicle safe, which makes it difficult in putting them all together.