LIDAR vs Radar vs Sonar: Which Is Better for Self-Driving Cars?14 min read28/05/2018
With remote sensing technologies like LIDAR, radar, and sonar, vehicle manufacturers empower self-driving cars to detect objects on the road. However, each of these systems has its working principle even though they have the same function. While sonars and radars are well-known devices that have been for many years, LIDAR is a relatively new technology. However, it comes with its drawbacks as well.
In his interview for The Verge, Elon Musk said that he saw the future of Tesla cars with radars, cameras, and sonars rather than LIDAR. On the other hand, Google’s Waymo self-driving car relies on laser- and radio-based technologies to track the road condition while Uber also equips their autonomous Volvo car with both LIDAR and radar systems.
Which system will most autonomous vehicles be equipped with in the close future? Our sonar vs LIDAR vs radar automotive comparison will sort things out.
LIDAR is a method of object detection, mapping, and range finding. While this term often appears in many sources as an acronym for Light Detection and Ranging, it has been originated from combining the words “light” and “radar”.
To detect objects on the surface as well as their size, form, and disposition, LIDAR scans environment by using laser light pulses.
It creates a three-dimensional digital copy of scanned environment by emitting near infrared, ultraviolet, or visible light. LIDAR can detect nearly any kind of material like rocks, plastic, metal, chemical compounds, and clouds.
Since this laser-based technology has high accuracy, it can even detect rain and aerosols.
National Geographic uses LIDAR to find ancient Maya graves, replicate buildings in a digital form, and generate a comprehensive geometry of urban spaces in a richly elaborated 3D model.
The LIDAR system for autonomous cars typically consists of four major components:
- Laser: a device that emits light waves of the 600-1000 nm length. This value makes a laser eye-safe while ensuring a suitable object detection accuracy.
- Scanning optics: this technology stack consists of sensors and dual oscillating plane mirrors. It points the light at the necessary direction to generate a digital copy of a certain piece of environment.
- Receiver electronics: this component captures reflected light waves. Capturing reflected light allows the LIDAR system technology to understand a form, size, speed, and distance to a certain object.
- Navigation system: self-driving vehicles typically use a Global Positioning System (GPS) to determine both disposition and orientation of the sensor.
How Waymo uses a LIDAR
Waymo uses the LIDAR system as a means of the main sensor. Mounted on the roof, it replicates a digital copy of the surrounding.
Because of the mapping capability of the laser-based technology, it can detect any obstacle on the road as well as other vehicles and pedestrians. Furthermore, the car can determine the direction these pedestrians face thus ensuring accurate prediction where they will walk next.
Besides LIDAR, Waymo has nine cameras to track the road condition: one forward facing camera stationed on the roof and eight others installed around the car. These cameras track the environment in 360 degrees.
In addition to lasers and cameras, Google self-driving car has a radar system that can detect objects in fog, snow, or fog.
Radar stands for Radio Detection and Ranging. This system uses radio waves to detect objects in the environment. Radars can determine the distance to a certain object as well as its speed and exact disposition.
The radar system technology uses special antennas to emit radio waves. It can detect objects made of solid materials including most metals.
The aviation industry widely uses radars. Equipped with these object detection systems, planes have been scanning the air since 1938. Military fighter aircraft typically use air-to-air targeting radars to identify other air fighters.
The radar technology stack typically consists of the following components:
- Transmitter: a device that emits radio signals in predetermined directions.
- Waveguide: a component that connects a transmitter to an antenna.
- Duplexer: a switch between a transmitter and antenna.
- Receiver: a component that captures reflected radio waves.
- Processing and displaying system: electronic components that transform radio signal into a human-readable format.
How Uber uses a radar
Unlike Waymo’s car, Uber’s self-driving vehicle has a front-mounted radar. It provides a 360-degree view. This radar is aimed at detecting other vehicles and large obstacles on the road.
Radio waves aren’t capable of recognizing the signature of a person. However, it’s likely to at least identify a pedestrian as a moving obstacle thus confirming data from the LIDAR mounted on the roof.
In addition, Uber’s autonomous car has short- and long-range optical cameras stationed around the vehicle. They identify road signs in real-time using special smart algorithms.
Sonar stands for sound navigation and ranging. This is an object detection technology that uses sound waves to detect objects in the environment. Among a LIDAR, radar, and sonar, the latter system is the oldest.
Self-driving vehicles like Tesla use active sonars which, unlike passive ones, both emits and receives reflected sound echo. Ultrasonic systems are aimed at detecting large objects made of solid materials.
The marine corps use sonars stationed on ships and boats to detect submarines in the water. The reason why the navy doesn’t apply radars and sonars for these purposes is in that water reflects both light and radio waves.
High-level sonars with narrow acoustic-wave beams can detect small boats as well as fish and frogmen in the deep water.
The ultrasonic system typically consists of the following components:
- Acoustic pulse generator: a device that emits soundwaves
- Transducer: a component that allows a transmitter to emit acoustic waves in narrow beams.
- Acoustic pickup: a receiver for capturing reflected sound waves.
- Amplifiers: electronic devices that increase the amplitude of sound waves.
- Delay timer: a component that calculates the echo delay. This data allows the ultrasonic software system to determine the distance to detected objects.
- Indicating display: A screen that presents processed data in a human-readable format.
How Tesla uses a sonar
In their Model S, Tesla uses the ultrasonic system based on 12 sensors stationed around the car. These ultrasonic sensors ensure a 360-degree view and can detect nearby obstacles like a vehicle, child or dog.
Tesla’s sonar system is functional at any speed regardless of whether the autopilot is active or not. This feature helps a driver detect objects in blind spots.
In addition to a sonar, Model S has a forward-looking radar. It can track the road up to 500 feet ahead of the self-driving car.
Similar to other manufacturers of autonomous vehicles, Tesla’s vision system has a forward-facing camera on the windshield of the car that functions as a backup to the radar and can identify road signs and traffic lights.
Tesla’s autopilot system also involves the GPS system that tracks the car’s position on the road and special technology stack that enables the car to automatically change lanes when possible and needed.
Comparison of object detection systems
There’s a set of differences between sonar vs LIDAR vs radar remote sensing. Each of these systems has its advantages and disadvantages. In this paragraph, we will cover why car manufacturers equip their autonomous vehicles with certain systems.
Furthermore, we will compare these technologies by their working principle, accuracy, deceiving methods, data update speed, cost, and size to determine which of these systems is better for automotive.
The working principle of the object detection technologies is pretty similar. They all use signal reflection as a way to detect objects as well as determine their speed and distance to them.
An algorithm of the remote sensing system functioning:
- A transmitter emits a signal.
- The signal reaches an object.
- The signal reflects from the object and returns.
- A receiver captures the reflected signal.
The difference between LIDAR and radar as well as sonar lies in the type of signal they use. A LIDAR emits light pulses, radar transmits radio waves, and sonar uses sound echo.
Since these systems are all based on the signal reflection principle, they determine the distance to a certain object as well as its speed in a similar way.
Each system determines how far this object is by measuring how much time it takes for a signal to reach the obstacle and return to the receiver.
Since the speed of the signal is known (the speed of LIDAR and radar signals equals the speed of light; the speed of a sonar signal equals the speed of sound), it’s easy to calculate the distance to this object using a simple formula:
S=V*t, where S is for distance, V is for speed, and t is for time.
When the object is moving, any of these remote sensing systems uses the echo-Doppler shift method to determine the speed of this object by measuring a pitch of the echo.
When it comes to the sonar vs lidar vs radar accuracy, we should better dive deeper into the characteristics of signals of each system to determine which one is more accurate.
The length and frequency of light, radio, and sound waves:
- Ultrasonic waves: 1.5-15 km; 20kHZ-200kHz.
- Radio waves: 1mm -100 km; 300 GHz-3 kHz.
- Visible light waves: 400-700 nanometers; 430–750 terahertz.
These values help us understand which system can “see” further than others and which one is the most accurate. The longer the wavelength is, the further a system can detect an object. On the other hand, the lower the frequency is, the more accurate the system is.
Radio waves can travel much further than light or sound. Therefore, a self-driving car equipped with a radar can detect other vehicle running 60 miles away.
The thing is though, whether autonomous cars really need to be capable of detecting objects that far since it’s usually enough a couple of miles to avoid an accident. Thus, ultrasonic systems seem to be more suitable.
In fact, car manufacturers equip their self-driving vehicles with sensors that track objects less than a half of a kilometer no matter which type of waves they use.
One way or another, the high-frequency signal ensures a high accuracy. Therefore, in the sonar vs radar vs LIDAR self-driving, this characteristic makes the laser-based system the most accurate remote sensing technology.
To determine the most reliable object detection technology, we will consider how easily each system can be deceived.
How to deceive a LIDAR
- To deceive a laser-based system, you can use two transmitters that have the same signal wavelength as the original LIDAR has. By creating fake echoes with these transmitters, you can make the system see objects further or closer than they really are.
- To spoof a LIDAR, you can use a transceiver and two pulse generators to create a number of laser pulse copies. When the sensor built-in receiver captures the signal, it will “detect” an object that doesn’t exist.
To jam a police LIDAR, car drivers use special devices that create light noise. Equipped with a receiver and transmitter, jammers capture light pulses sent by a LIDAR and then emit a signal with the same wavelength and higher intensity. Thus, a police LIDAR either determines the car speeding at zero miles per hour or can’t detect this vehicle at all.
The main drawback of laser-based systems is that light reflects from fog, snow, or raindrops. LIDARs can work properly only under certain weather conditions. The military aviation industry uses this light property to protect aircraft from laser-guided missiles.
Equipped with the infrared missile attack countermeasure, some aircraft apply decoy flares to make a laser-guided missile search for the heat signature from the flare rather than the aircraft’s engine.
How to deceive a radar
- To make a radar determine the speed of a running vehicle as zero, hackers generate a radio noise using a radio wave receiver and transmitter with the same wavelength as the vehicle-based radar has.
- Another way to deceive a radar is to make radio waves reflect in the wrong direction from a certain object or even miss it using the right shape design for the object.
Radio beams reflect well from solid vertical surfaces. Composed of a wide range of curved surfaces across its airframe, the American stealth aircraft B-2 Spirit deflects radar beams to stay undetected. Its flying wing design looks like an infinite flat plate.
In addition, a rear part of the B-2’s flying wing has an angular form that allows the aircraft to deflect back radio waves.
How to deceive a sonar
Sonar deceiving methods include signal jamming similar to the one used for deceiving both LIDARs and radar. The only difference is in using the right type of a signal.
When it comes to radar vs sonar reliability, the signal reflection of both systems depends on the form of an object they target.
Navy uses this ultrasonic quality to keep submarines undetected for a sonar. Because of the special coating made from a phononic crystal, a submarine hull can curve sound waves thus making them bounce over and over again.
Moths have been using this method for ages to deceive bat echolocation. Their tails whirl in circles to bounce back bat’s sonar pings. Thus, moths remain undetected for bats.
Therefore, within the sonar vs RADAR vs LIDAR autonomous driving comparison, a radar and sonar are more predictable. Despite all these systems are deceivable, the dependency on weather conditions makes a LIDAR the least preferable system when it comes to reliability.
Data update speed
How fast data gets updated in remote sensing systems depend on how fast a signal can reach an obstacle and return to receiver.
When lightning appears – at first, we see it. Only later, in a few seconds, we hear the sound.
Since the light is 1,000,000 faster than the sound, both laser pulses and radio waves reach obstacles much faster than sound waves do. That’s why LIDARS and radars detect objects on the road faster than sonars do.
Furthermore, a high speed of radio and laser signals allows a radar and LIDAR to track the position of a moving object in a more accurate manner than a sonar.
That’s why the latter sensor is mostly used as a means of a rear obstacle detection system. When parking, a driver mostly monitors either already parked vehicles or other slowly approaching cars.
The first public autonomous vehicle prototype built by Alphabet (Google) on a base of Toyota Prius used a LIDAR called Velodyne HDL-64E. This device used a set of 64 sensors and could see in 360 degrees. Its price was about $75,000.
In total, extra equipment of Google’s driverless cars cost about $150,000. However, according to Bloomberg, Google managed to significantly reduce the cost of their LIDAR in five years. Their new laser sensor costs about $7,500 which still is pretty expensive.
Although, a Silicon Valley-based startup called Luminar has invented a LIDAR that costs only $3 per unit. Even though their sensor hasn’t got a wide coverage yet, the company has already partnered with Toyota.
However, radars and sonars are relatively cheap. On Amazon, you can find either a radar or sonar for $30-$400.
As for now, LIDARs are too expensive to make driverless cars affordable. When it comes to the comparison of sonar vs radar vs LIDAR self-driving vehicle system cost, radio- and echolocators are more preferable.
A LIDAR for self-driving cars looks like a coffee-can-size lamp while radars usually have a size of a fist. When it comes to sonars, their size is even smaller since manufacturers embed them into bumpers.
Furthermore, LIDARs need to be mounted on top of the car as high as possible to be able to detect objects around the car, so Waymo has a laser sensor mounted on a riser that is installed on the car roof.
NXP, a chipmaker from the Netherlands, created a 0.3-inch size radar transceiver in 2016. Perhaps, such transceivers will soon become a part of sensing systems of most autonomous cars.
As of now, the size of a sonar makes it the most compact sensing system.
|Type of a signal||Light pulses||Radio waves||Sound waves|
|Data update speed||✔||✔|
While radiolocation and ultrasonic systems are cheap and compact, lasers can accurately predict a vehicle’s trajectory. While a sonar and radar are good at detecting objects at long distances and both of them work well under nearly any weather condition, a LIDAR provides more accurate data and generate a 3D image of surroundings which allows software to properly determine whether an object is a vehicle or pedestrian.
Our comparison shows that there’s no perfect sensing system for self-driving cars since each solution has its drawbacks and benefits.
That’s why self-driving carі use different sensing systems instead of relying on one specific solution. Such approach minimizes the error rate and provides accurate data under nearly any condition.
Which system will future autonomous cars use? Even if we assume that any of these three types of sensors will cost equally, radars and sonars are more likely to be a part of future self-driving cars than LIDAR.
LIDARs are bulky and they make any self-driving vehicle look like a police car with a siren on top. Moreover, today’s software can clearly distinguish pedestrians from vehicles using data received from camera sensors.