Sensors in Self-Driving Cars

The part of Self-Driving Cars that no one ever talks about

6 min readOct 28, 2019

Whenever you say “Self-Driving Cars”, people automatically think “AI”, “machine learning”, “back to the future” or “Elon Musk”. But the one thing that no one ever thinks of is: “Sensors”. Yeah, I know you probably weren’t expecting that (the title maybe gave it away).

LIDAR and RADAR are just as important as all the fancy algorithms are to a self-driving car, just that people don't want to care as much about the nuts and bolts behind self-driving cars. By the end of this article, I guarantee that you will appreciate tech like LIDAR and RADAR just as much as you love those CNN’s that you write in PyTorch.

Audi Aicon, Self Driving Concept (it looks like a BEAST!!!)

Without sensors, a self-driving car is useless. It needs sensors to function. A self-driving car uses many different types of sensors to allow itself to perform the necessary tasks to drive autonomously. When all the sensors combine their data, it allows the car to thoroughly understand its environment. These sensors include LIDAR, RADAR and ultrasonic sensors.

Lidar

LIDAR (Light Detection and Ranging) uses laser pulses to create 3D maps of the car’s environment. Its main purpose is to create highly sophisticated 3D maps of the car’s environment, but it can also determine how far objects are from the car. Due to these features of LIDAR, it has a huge use for engineers in the AV space.

A LIDAR unit is made up of a

GPS: Records location of the car
IMU: Measures angle of the scanner relative to the ground
Clock: Records time that the laser pulse takes to return to the scanner
Laser: Intense beam of light that can travel for long distances without spreading

A LIDAR unit can create these maps by firing millions of laser beams every second. It then calculates how long it takes the laser beam to return to the car. By firing millions of these beams, the LIDAR system can create a 3D image of the car’s surroundings (as long as the object is in the line of sight).

Like I said earlier, the goal of LIDAR in self-driving cars is to create a detailed, 3D picture of its nearby surroundings. This can be effectively done through LIDAR’s lower Field of Vision (>200m radius), and its high precision and detail.

LIDAR may be an awesome technology, but it has its downfalls. First of all, it’s really expensive. By that, I mean $75000 per unit, which will make self-driving vehicles out of the reach of the general public. However, many startups are working to decrease this cost.

One of these startups is Luminar, who created a LIDAR unit that only costs $1000. Waymo, Google’s self-driving car company, also started selling their LIDAR systems for less than $7500 a unit. The price of LIDAR will eventually decrease, it’s only a matter of time.

Another downfall of LIDAR is that it does not look pretty. It looks like a jumble of lasers and wires thrown together on the roof of a self-driving car. I’m not sure about you, but I would not want my car to have that on its roof.

Let’s be honest, that LIDAR is not pretty

Overall, LIDAR is an awesome technology that has hugely benefited the autonomous vehicle industry in the past decade, but it still needs some further innovation before it is consumer-ready.

Radar

Radar (Radio Detection and Ranging) uses radio waves to create a picture of the car’s surroundings. Since radio waves are capable of travelling further distances, RADAR is capable of creating larger maps than the LIDAR used in self-driving cars.

RADAR is a much older technology than LIDAR. RADAR was invented in 1935 by the British physicist Sir Robert Wattson-Watt. Over the years, RADAR was commonly used by planes and ships to navigate at night, and also in police speeding guns to check which cars are speeding. Now they are making their way into the realm of cars.

However, RADAR in cars is not a new thing. Tech like cruise control, park assist and blind-spot detection have been out for over a decade. At the time (the late 90s), the commonly used 24GHz RADAR was getting old and had slow bandwidth. However, lots of research was conducted with the 77GHz RADAR, which allowed tech like Adaptive Cruise Control and Cross Traffic Alert to become a reality.

Today, there are 3 types of RADARS used in cars: Short Range, Medium Range and Long Range Radar. Self-driving cars use all 3 types of RADARs which each have their own unique functionalities. Here are some of their uses in self-driving cars.

Short Range Radar

Park Assist: Self-parking feature that is common on many of the newer cars today.
Blind Spot Detection: Alerts the driver if there is any object in the blind spot
Object Detection: Detects objects which are close to the car (ex. Pedestrians, cyclists, other cars)

Medium Range Radar

Rear Collision Warning: Alerts the driver if the car thinks that it may crash into something behind it.
Cross-Traffic Alert: If the driver tries to switch lanes, but there is another car in that lane, the side mirror will light up, alerting the driver not to cross lanes.
Object Detection: Detects objects which are in a close range to the car (ex. Traffic lights, road signs):

Long-Range Radar

Adaptive Cruise Control: Car adapts its speed based on the speed of the car in front of it
Forward Collision Alert: Alerts the driver if the car thinks that it may crash into something in front of it.
Object detection: Detects objects within a long range of the car (buildings, other cars, etc)

RADAR is also used inside the car to check if the driver is still paying attention to the road.

Ultrasonic Sensors

Ultrasonic Sensors use SONAR (Sound Navigation and Ranging) to help drivers with tasks like parking and nearby object detection. Ultrasonic sensors function similar to LIDAR and RADAR but use sound waves instead of laser or radio waves.

Ultrasonic sensors are great because they can see through the car’s line of sight, which means that it can see objects that are hidden behind another object which is in the car’s line of sight. They’re also really cheap and work great in bad weather; like hail, thunder, snow, fog and at night. Some newer models also have recognition capabilities close to LIDAR, however within a much smaller range.

Unfortunately, ultrasonic sensors have a much smaller range than LIDAR or RADAR (<10m). They also cannot detect objects at high speeds, due to sound waves being slower than laser pulses (used in LIDAR) and radio waves (used in RADAR).

Key Takeaways

Self-driving cars need sensors like LIDAR and RADAR to drive themselves
LIDAR uses laser pulses to create detailed 3D maps of its surroundings
LIDAR is very expensive and ugly😂, two things that hopefully will improve with time
RADAR is similar to LIDAR, except it uses radio waves.
There are Short, Medium and Long Range RADAR systems which each have their contribution to the self-driving car.
Ultrasonic (SONAR) is similar to RADAR and LIDAR, but it uses sound waves
Ultrasonic sensors are good for nearby object detection at low speeds but struggle with longer ranges (>10m) and at high speeds.