Lidar is an acronym for “light detection and ranging.” It is sometimes called “laser scanning” or “3D scanning.” The technology uses eye-safe laser beams to create a 3D representation of the surveyed environment. Lidar is used in many industries, including automotive, trucking, UAV/drones, industrial, mapping, and others.
A typical lidar sensor emits pulsed light waves from a laser into the environment. These pulses bounce off surrounding objects and return to the sensor. The sensor uses the time it took for each pulse to return to the sensor to calculate the distance it traveled. Repeating this process millions of times per second creates a real-time 3D map of the environment. An onboard computer can utilize this 3D map of the surrounding environment for navigation.
Yes, laser and lidar technology has been used safely for a long time in a wide variety of applications worldwide. Lasers are used today in the public in supermarkets, light shows, and home security systems. Manufacturers follow strict guidelines to ensure the sensors pass eye-safety standards.
Yes, Velodyne lidar sensors meet the FDA eye-safety standards under IEC 60825.
Lidar sensors do not drive a car. They provide the 3D vision for the car’s computer and mechanical system to make driving decisions.
Lidar sensors make autonomous vehicles possible by providing a high-resolution, real-time 3D view of the surroundings. Lidar provides autonomous vehicles 3D vision by generating and measuring billions of data points in real time, creating a precise map of the ever-changing surroundings for the vehicle to safely navigate. Lidar’s exceptional distance accuracy allows the vehicle’s system to identify and avoid objects. With lidar, the vehicle can “see” up to 300 meters on all sides with accuracy within a few centimeters. This enables the vehicle to drive itself at high speeds and navigate the road safely. Velodyne is the original inventor and market-leader in 3D lidar technology. Velodyne currently supplies more than 40 OEMs and new tech entrants worldwide. As a result, Velodyne’s sensors have been tested, validated, and utilized over millions of real-world road miles.
No. Velodyne is unique in offering a multi-generational family of lidar sensors with 360-degree Field of View (FoV). There are certain applications, such as Advanced Driver Assistance Systems (ADAS), that do not require 360-degree FoV. In those cases, lidar with smaller horizontal FoV are available, such as Velodyne’s Velarray sensor, which has a horizontal FoV of 120 degrees.
Some people use the term “solid state” for lidar such as the Velarray, when in fact this technology is better described as having a Limited Field of View (LFoV). LFoV lidar are made with moving parts, such as moving mirrors. In contrast, 360-degree lidar are made of an internal solid-state mechanism that spins. The spinning mechanism is based on ball bearings, such as those used in robust jet engine systems. Frost and Sullivan describe Velodyne’s spinning lidar technology as solid-state hybrid (SSH) lidar.
There is no inherent difference in 3D vision quality between the 360-degree lidar and the LFoV lidar. However, the LFoV provides a slice of the environment. Its limited view typically spans from 90 degrees to 120 degrees. To “see” 360 degrees of the environment, LFoV pictures must be stitched together by the car’s computer. This requires extra processing and problems can arise in the “stitch lines.”
All sensors have defined field of views (FoVs) and resolutions. Blind spots or gaps in detection might occur based on the combination of sensors placed around the vehicle at specific locations. When multiple lidar sensors are mounted with care, placed to create overlapping FoVs around the vehicle, it is possible to minimize or eliminate crucial blind spots.
Sensors such as radar and cameras have blind spots. Cameras have nighttime blind spots with limited visibility in high glare or nighttime conditions. Radar has multiple blind spots because its technology is not refined enough to accurately identify objects.
Depending on the application and the associated field of view, lidar with 8 to 128 laser beams (or lines) can be adequate. For example, a forklift can use a lidar with fewer sensor channels because it moves more slowly than a car. A car can be equipped with a 32 channel lidar to drive autonomously up to 35 MPH. To drive autonomously at 65 MPH, the driving system should be equipped with a lidar containing 128 channels. As a robot or vehicle moves faster, the sensor requires more channels to produce images in high resolution.
Cameras produce 2D images of the environment and companies are installing them around vehicles to use for navigation. However, there are serious problems with the accuracy of proposed camera-centric systems which have not been solved and will likely not be solved soon.
Lidar “sees” in 3D, a huge advantage when accuracy and precision is paramount. The laser-based technology produces real-time high-resolution 3D maps, or point clouds, of the surroundings, demonstrating a level of distance accuracy that is unmatched by cameras, even ones with stereo vision. Wheras cameras have to make assumptions about an object’s distance, lidar produces and provides exact measurements. For this reason, autonomous or highly automated systems require lidar for safe navigation. The ability to “see” in 3D can’t be underestimated. Lidar produces billions of data points at nearly the speed of light. Each point provides a precise measurement of the environment. Compared to camera systems, lidar’s ability to “see” by way of precise mathematical measurements, decreases the chance of feeding false information from the vision systems to the car’s computer.
Camera performance is also greatly impacted by environmental conditions (e.g., bright sunlight or glare, darkness) and is therefore more susceptible to unpredictable blind spots and generating false positives or negatives. Whereas cameras are dependent on ambient light conditions and face challenges with darkness and glare, lidar provides its own light source and can therefore “see” in all lighting conditions.
Lidar has an additional technological advantage over camera systems: lidar allows the vehicle’s computer to “see” the driving environment from an overhead, bird’s eye perspective. The car navigates not only from a traditional driver’s point of view, but can also “see” itself from the perspective of a bird flying overhead, similar to the views offered in many video games. Thus, lidar “sees” more comprehensively than a person, simultaneously looking down on all sides of the car, road, and traffic.
With accuracy and safety in mind, a lidar-centric autonomous system is a necessity. This means lidar is the central sensor and small inexpensive cameras can be added to the system for redundancy and extra care.
Autonomous and highly automated systems should be lidar-centric, but other sensors such as cameras and radar, can be helpful. Lidar should be the foundation of the system, providing the most comprehensive performance across a wide range of road conditions. Lidar can be supplemented by inexpensive cameras and radar for redundancy and extra care.
Higher resolution lidar sensors generate billions more data points, providing the system with even more information upon which to base driving decisions. High-resolution lidar imaging is critical for enabling accurate object detection and classification, helping ensure that the vehicle drives safely and avoids collision. Resolution becomes even more important as the speed of the vehicle increases. To maximize safety as speed increases, which require the vehicle to detect and classify objects at a greater distance to maximize safety.
Lidar resolution and object detection are enhanced by the very movement of a vehicle. In other words, the laser and data points that fall on an object multiply as a car moves down the road, filling in the details of the image. These allow the vehicle’s system to identify positions and gestures of pedestrians, traffic signs, even leaves on a tree. However, the required resolution for object detection increases as a vehicle speeds up. Since there is less time for the process, more lidar channels are needed. As the vehicle approaches highway speeds high resolution is important to accurately detect and classify objects, especially at a distance. A car moving at 65 mph might have only a few seconds to slow down and prevent a collision with an object in its path 200 meters away. Higher resolution enables the car to more accurately identify and classify the distant object early on and avoid collision.
Current lidar sensors typically use one of two wavelengths, either ~905 nanometers (nm) or ~1550 nm. We can better understand the differences between these wavelengths by comparing their performance with regards to safety, water absorption, and power consumption:
Sensors using 905 nm and 1550 nm wavelengths achieve eye-safety certification via compliance with the FDA eye-safety standard IEC 60825. If sensors are designed to meet eye-safety standards, both wavelengths can be used safely.
Given the variety of weather conditions cars encounter on the road, how a sensor’s laser pulses interact with water is an essential issue for automotive lidar. A key difference between 905 nm and 1550 nm wavelengths is that 1550 nm waves get absorbed by water to a much greater extent than 905 nm waves, as discussed in a research paper published in Opto-Electronics Review. As a result, 1550 nm waves get substantially weakened under conditions of rain, fog or snow compared to 905 nm waves.
To offset degradation due to water absorption and to achieve high range, 1550 nm systems generally send out more laser light to achieve performance comparable to 905 nm systems. As a result, 1550 nm systems typically consume more electrical power. Higher power could limit the maximum achievable operating temperature due to the challenge of dissipating the extra heat that is generated using higher power. A high-power requirement could also create the need for larger, more extensive equipment which must be stored in the vehicle. In general, the lower the lidar sensor’s power consumption, the more energy that an autonomous vehicle system can devote to other driving functions, such as object detection and avoidance.