Let’s dig into one of the most common SLAM (Simultaneous localization and mapping) navigation tools and see what assets and liabilities come with it.
The most popular SLAM systems use optical sensors and among the top ones is LiDAR-based that uses 2D or 3D LiDAR scanners. This is what we’re going to put under a microscope in this article.
LiDAR is a method for distance measurement. The name stands for Light Detection and Ranging. Basically, it’s a version of radar but instead of radio waves, it uses light. That would be the most basic explanation. Read on to know the details.
The LiDAR device consists of a range measurement sensor that repeatedly emits a pulse of light. After the light hits a target, it is reflected back to the sensor which then determines the distance to the object by measuring how long the light needs to meet the target and return.
In addition, the range measurement sensor is placed on a rotating platform which enables the device to take readings at multiple points within 360 degrees. As the sensor spins, range measurements are taken quickly (up to about 10000 samples per second), providing a two-dimensional view of the entire robot’s surroundings.
The result of a 360-degree view sweep along with taking multiple range samples is a raw map. The next step in this process is to take the 360-degree scans and gather them into a more complete map. When the robot moves in its environment, it can determine where it is in relation to the present and previously scanned data (the localization process) and then, performs new scans and adds them to the map. This is where the process’ name came from, that is, “Simultaneous Localization and Mapping” (SLAM).
Combined with mechanisms that move the laser in a plane or another pattern, users are provided with a series of distance readings at different angles and/or directions. Based on this 2D or 3D range data, users can convert radial coordinates to Cartesian ones and, as a result, create a 2D or 3D contour map of the scene.
Long story short – the way most LiDARs work is calculating the distance to an object by illuminating it with a number of transceivers. Each transceiver rapidly sends pulsed light and measures the reflected pulses in order to determine distance and position.
Now that we’ve covered the basics, let’s find out what some pros and cons of LiDAR are.
The incredible speed at which light travels requires highly accurate laser operation to track the distance from the robot to each target with utmost precision. That’s what makes LiDAR perfectly suited for both fast and accurate scanning (accuracy of 0.15-0.25 m). But, it only applies to what it is able to see. One of the major cons of 2D LiDAR is that if one object is obscured by another at LiDAR height, or an object has an inconsistent shape that doesn’t have the same width throughout the body, this information will be lost.
A considerable advantage of the LiDAR device is its independence of lighting conditions and weather, except for fog and heavy cloud cover. Unfortunately, the device also absorbs laser pulses by water, asphalt, and tar. The absorption and dispersion of LiDAR waves by atmospheric aerosols is in turn an advantage from the meteorological point of view.
Also, LiDAR one-ups structured light techniques such as triangulation or stereo vision as it can read data from very long distances without the need for a big baseline between cameras or emitter/detector pairs, which can be significant for smaller vehicles.
Some of the LiDARs are able to specifically detect reflectors in laser scanning, delivering a low computational power triangulation system for localization. However, the use of reflectors and triangulation comes with the complication of having to see three or more reflectors at any point while the robot is moving. It is challenging when the operating area is divided by some shelving, machinery, and other objects.
A large volume of data sets is another LiDAR drawback.
Each SLAM algorithm is conclusively based on the environment sensor readings. When no objects are in the correct position for the sensors to read them, the implementation of camera-based 3D LiDAR detection or 3D stereo depth sensing can significantly increase localization efficiency. These sensors, however, entail being potentially more expensive and having much higher computational requirements. There are low-cost LiDAR devices too though. An example of such a cheaper device is RPLidar A2 which is one of the three most popular add-ons used with Leo Rover (check out the other two in our article).
LiDAR, being among the top tools for SLAM navigation, provides its users a lot. But we all know nothing is perfect and it’s no different in this case. As shown above, alongside multiple possibilities of the device, some drawbacks can also be found. Do they undermine the LiDAR effectiveness, though? You're the one who needs to render the verdict. If you decide to use LiDAR in your application, see how to connect it to Leo Rover.