LiDARs have grown in popularity in recent years as better manufacturing techniques and mass production have lowered the price of this technology.
But since no single LiDAR solution is best for all applications, choosing the right LDAR can be tricky, a >$50k LiDAR is great for autonomous vehicles but it may not be able to properly detect an obstacle less than a meter away, a <$1k LiDAR can do that job effectively but may not be able to look very far.
If you are considering using a LiDAR for any of your projects keep reading to find out how to choose the right LiDAR for your application.
What is a LiDAR used for?
LiDARs have been used in surveying, forestry, geology, mining, architecture, and mapping for many years, but they are now becoming a popular component of autonomous vehicle and robot applications.
In architecture, for example, you can get a handheld LiDAR device that you carry while walking around a building. The device creates a 3D point cloud, so you can match the scan of the building to the CAD file of the construction design to see if the real building measures up and also to assist in the as-built process and survey.
In mining, you can use LiDAR for safety and surveillance. Instead of putting humans in danger, you can send a LiDAR-equipped robot down mines to detect anomalies like cracks in the walls or other physical objects that may be a hazard. Although you could do this with cameras as well, cameras aren’t great at picking up depth, even if you can see a crack in the wall, a camera can't tell you how deep it is. Cameras are also heavily influenced by lighting conditions, which may not be ideal in some environments.
Long range vs. short & close range
First, you have to look at the LiDAR range. As a rule of thumb, long-ranging LiDARs with a range of 100m or more are more expensive.
For example, long range LiDARs work work well for navigating vehicles that are moving quickly., but they usually only detect objects at mid- to long-range; they don’t work at close range, such as for detecting objects within 1 m.
If your application involves a fast-moving vehicle, long-range LiDARs can help , because you have to detect objects from far away and take necessary actions before the objects come closer.
Indoors or with slower-moving vehicles, you typically don’t need long-range. If you’re operating a slow moving robot you typically need to see up to 25m, you can work with a short-range LiDAR.
Close-range LiDARs are used to detect objects in close vicinity. For example you may need a close range LiDAR on a vehicle to detect obstacles while opening the doors. Usually you would combine a long or short range LiDAR with a close range one.
2D vs. 3D
2D LiDARs will detect an object. 3D LiDARs will produce a more detailed point cloud which can be used to determine the shape and depth of the object, and determine what it is, rather than just telling the robot or vehicle “there’s something there.” 3D gives you better understanding of the surroundings, so your robot or vehicle can distinguish between a tree, a sign on the highway, or an oncoming truck.
The difference is really in the information that can be extracted from them. For instance, using them in a building, 2D LiDARs can only give you a floor plan whereas, with 3D LiDAR, you can get the 3D map and location and height of every object in the building. So which one to use depends on the application.
If you are navigating in a flat environment and your robot is not tall, perhaps a 2D LiDAR is enough. However, for 3D outdoor environments where there are obstacles like hanging tree branches or pipes, you need 3D perception. 3D LiDARs are also useful in other applications like terrain classification and segmentation.
More considerations for different applications
- Field of View (FoV): Field of View is a critical feature of any LiDAR. It is measured separately along the horizontal and vertical axis. An autonomous vehicle is usually equipped with one or more LiDARs with 360° horizontal FOV, or a larger number of directional LiDARs to ensure the 360° degree view is available. It may not need a larege vertical FoV, however, as it may only need to look at the road and perhaps 30 to 40 degrees above the road surface.
If your application is obstacle avoidance within a constrained space you may not need a 360 degree LiDAR. The 360 degree surround capability is not an asset in all cases. If you only need to look straight ahead a 360 degree LiDAR would give you extra and non-essential data and hence put unnecessary load on system processing.
- Number of channels: A LiDAR with more channels will produce a denser, more detailed point cloud. The amount you choose to get may boil down to your budget, processing capabilities, size of the objects, distance from the objects.
- Vertical separation between adjacent channels: If you are trying to detect small objects you need to make sure that the separation is as small as possible, otherwise there is a chance that the object may fall within the two adjacent channels and go undetected.
This can, however, be mitigated by moving the LiDAR so that the adjacent lasers fall on different points and are thus able to detect the object. Some LIDARs have smaller separation near the center of the vertical FOV and bigger separation away from the center. This helps detect objects near the focus of the LiDAR.
- Detecting Low Reflectivity Objects: The range and accuracy of LiDARs varies considerably depending on the object’s reflectivity or how much light is reflected back from the object. Known as Remission, it is expressed in percentage. A white dry wall would have close to 90% remission while something like coal would have around 5% remission.
A LiDAR’s range is typically expressed at 80% remission, it can go down to 1/4th or even less at 10% remission. If your project involves looking at or detecting darker objects or objects that do not reflect a lot of light you may want to ask the LiDAR manufacturer to provide you with the range and accuracy at different remissions… unless of course you are looking at a white drywall all the time!
- Points Per Second: PPS or Points Per Second is another good spec to look at. It shows how many laser measurements are being taken per second. More points will result in a more detailed, denser and accurate point cloud.
PPS is primarily the function of the frequency of the LiDAR in Hz, number of channels and number of returns. A LiDAR operating in dual return mode would provide double the points per second. Similarly a LiDAR operating at a higher frequency would yield higher PPS.
- Safety-certified LiDAR: If your robot or autonomous guided vehicle needs to operate amongst humans, you might want something which is safety-rated. There are some companies that offer safety-rated 2D LiDARs but this option is very limited at this point in time. The safety rated LiDAR would be just one component of the overall system and other considerations would need to be made to make the entire system safety certified.
- LiDAR or a Complete Solution? You might need a complete mapping solution, not just LiDAR. Your technical resources will determine how easy it is to put an extensive ecosystem into place. The LiDAR can provide you with a point cloud but to make sense out of it you may need localization, time stamping and other valuable data. you may have to consider complete solutions that consist of LiDARs, IMUs, GPS, Cameras and software. You may also need to do some kind of post processing to use the data in a meaningful way.