​The Role of LiDAR in Autonomous Driving

Autonomous vehicles need to visualize, understand, and react to the world around them to operate safely. The first step is accomplished by a variety of sensors, including LiDAR (Light Detection and Ranging). LiDAR has emerged as a particularly important technology for automakers interested in pushing the current limits towards fully autonomous driving.

What is LiDAR?

LiDAR stands for Light Detection and Ranging and is commonly used in remote sensing applications, including autonomous driving. LiDAR systems have three major components: a laser that emits light pulses, a detector to measure the time it takes to receive the reflected light pulse signals, and a processor that calculates the distances to and locations of surrounding objects based on that data. This information allows an onboard computer to create a point cloud, a detailed three-dimensional picture of the world around the vehicle that includes people, animals, street signs, road markings, obstacles, and other cars and trucks. The technology is sometimes referred to as time-of-flight (ToF) LiDAR because it measures the time between a laser beam being sent out and when the reflection returns to paint an electronic picture.

Automotive LiDAR allows a vehicle to visualize the scene in detail using the point cloud. Laser pulses are emitted in rapid succession for real-time coverage—think tens of thousands to hundreds of thousands of pulses per second. The type of information collected is critical for the safe operation of autonomous vehicles, which is why the development and testing of LiDAR technology has become important to auto manufacturers pursuing smart cars.

Types of Automotive LiDAR

Several LiDAR technologies are relevant to self-driving cars. At the highest level LiDARs are differentiated as:

Mechanical LiDAR

Light pulses are emitted from a rotating assembly (e. g. on the roof), enabling complete coverage of the surrounding environment. While mechanical LiDAR has a 360-degree field-of-view (FoV), the device tends to be bulky and needs to sit on top of the vehicle. These constraints can make it impractical for commercial use outside of mapping or other specialized technological applications.

Solid-State LiDAR

The laser emitter and detector sit on a single microchip. A beam of light is emitted/received by the sensor without the use of moving parts. The tradeoff, however, is that you need multiple sensors to provide full coverage around a vehicle (limited FoV). The relative simplicity of this technology makes it the preferred option for many automakers. 

Here are some different approaches to solid-state LiDAR on autonomous vehicles:

• Flash LiDAR—A powerful laser emits a flash of light that illuminates its surroundings, sending reflected signals to an array of detectors. An advantage of this approach is that the sensor receives a wide view of the area from each pulse. 
• Frequency Modulated Continuous Wave (FMCW) LiDAR—While working under the same basic principle of emitting and receiving light pulses, FMCW LiDAR can also measure the frequency of the light waves. The processor then uses the Doppler Effect to determine how fast objects are moving towards or away from the vehicle. Other forms of solid-state lidar need to compare two signals to make this calculation, so they’re less efficient in this respect.
• Microelectromechanical system (MEMS) LiDAR—Rather than relying on additional rotating components to expand its field-of-view, MEMS LiDAR directs light pulses with many tiny mirrors that can move to cover a wider area. MEMS LiDAR is more compact than a pure mechanical rooftop unit but can be affected by the vibrations that accompany driving a vehicle since it isn’t strictly solid-state.
• Optical phase arrays (OPA) LiDAR—Multiple sensors are placed near one another and fire off in sequence, but the speed of the light passing through each lens is different. This difference enables engineers to curve the laser light, simulating rotation without the need for moving parts. 

The technologies around LiDAR are continuously evolving. Combinations, like adding mechanical components such as mirrors and/or prisms to solid-state based designs, are quite common today and form hybrid LiDAR systems.

Why is LiDAR Important for Autonomous Driving?

The purpose of automotive LiDAR is to use light to build a point cloud and help the car “perceive” its surroundings so onboard processors can safely navigate the environment without input from a human driver.

LiDAR is currently used in many Advanced Driver Assistance Systems (ADAS), like adaptive cruise control and lane keeping. While it’s not the only option to help autonomous vehicles visualize their surroundings, LiDAR has the following advantages:

• Speed—LiDAR can emit hundreds of thousands of light pulses per second for real-time coverage of the roadway, an essential part of making self-driving vehicles possible.
• Accuracy—The high resolution of the point cloud built by LiDAR sensors creates a precise 3D model of the scene around the car.
• Versatility—Automotive LiDAR operates in near-infrared or short-wave infrared wavelengths and works equally well in sunlight or at night. The laser pulses aren’t visible to humans.

The use of LiDAR on vehicles isn’t without its challenges, like system complexity leading to higher cost. LiDAR suppliers are hard at work to push the cost per unit down below  $1,000 and further on towards the $500 barrier – but still, at these price levels LiDARs remain more expensive compared to radars (below $100) and cameras (below $25). Additionally, the chips needed to build LiDAR systems can sometimes face supply chain challenges as manufacturers add more of these to their vehicles.

Weather can also create issues, as precipitation like rain and snow can reflect or absorb LiDAR signals. This absorption reduces the LiDAR range and impacts performance in inclement weather.

While there are no fully autonomous vehicles on the market today, LiDAR is likely to be along for the ride when they debut.

LiDAR versus Radar

While both technologies can detect objects, LiDAR and radar have major differences. LiDAR emits laser beams while radar uses radio waves. LiDAR’s advantages include higher resolution return signals for more detailed maps, while radar has a greater range. LiDAR provides 3D maps (point cloud) while radars typically operate in 2D (excluding developments around 4D imaging radar). Radar sensors are also currently more affordable than LiDAR units.

The genius lays in combining the capabilities of both technologies. LiDAR and radar can work in concert for ADAS and autonomous driving features and complement each other. This type of collaboration is referred to as “sensor fusion” and takes advantage of the strengths of each technology.

^{Figure 1: Comparison of LiDAR and Radar Sensors}

NI’s LiDAR Test Solutions

NI hardware and software support LiDAR testing for original equipment manufacturers (OEMs), suppliers, and researchers. Sensors can be tested at the individual level and as part of an integrated system. The ADAS HIL Sensor Fusion Test Workbench from NI partner Konrad Technologies allows engineers to connect and test a variety of sensors to simulate real-world operation in a vehicle.

With the proliferation of sensors needed to make autonomous driving a reality, many automakers’ test needs overlap those of semiconductor manufacturers. NI’s expertise and solutions span both sectors and can help companies shepherd sensor technologies through development and test and into production.

Next Steps:

• Learn more about NI’s approach to ADAS and Autonomous Driving Sensors and Electronics Test
• Get details about prototyping and testing FMCW Lidar