Introduction
Achieving accurate SLAM results with mobile laser scanners requires an understanding of the conditions that can affect scan quality. This document describes the three main challenges operators face in the field and offers practical tips for minimizing their impact.
Dense vegetation,
Moving objects
Lack of geometric features
Dense Vegetation
Vegetation poses challenges due to its irregular shapes and constant slight movements caused by the wind. This is detected by the device as noise. If possible, try to plan your scan to avoid these areas where the operator is surrounded with thick dense vegetation.
As this is not always possible, you can try to maximize the geometry of these areas, for example, put the device’s shipping box or backpack at the end of a narrow bushy path. You might also consider shifting more to the side so that the front laser does not directly "detect" the bushes. Additionally, traversing thick grass complicates accurate z-axis estimation, and we occasionally notice drift in the z-axis.
Moving Objects
Moving objects such as pedestrians or vehicles, especially large ones, can introduce uncertainties as they look like buildings that move, the device is not sure if the environment is moving or the operator is moving. The same issues occurs when doors open during a scan. Always try to open the doors before the scan begins or have another person open them before the operator comes to that room.
Lack of geometric features
If you are scanning expansive spaces lacking distinctive shapes, the device's lasers find no objects to detect and can not determine the device's movement. Similarly, in a long featureless corridor or tunnel, there is nothing to determine the position of the operator along the tunnel shape, so it might "slide" along the axis of the tunnel/corridor. You can always modify the scene with the device’s shipping box, traffic cones etc.
Loop closures
Remember that loop closures are only implemented at the processing stage. They are not activated during the scanning so it is common to see more drift during the scanning which can be fixed later at the processing stage.
The most effective way to define a loop closure is by revisiting areas that have already been scanned after a period of not encountering those locations. To exemplify this with a simple real-life scenario, picture yourself as a tourist in an unfamiliar city. You might begin your exploration at the central square and meander through the city streets "without a specific direction" until you inadvertently return to the same square where your journey began, after an hour of walking. You will recognize the location and subsequently be able to "reconstruct" your route in your mind, as the paths you took now connect.
In a scanning context, loop closure can be walking a full loop around a building and ending up where you started from. The processing software will recognize this area and be able to correct any accumulated error along the way. However, walking in small circular pattern is not considered loop closure as the lasers constantly see the same area.
For more information on Loop Closures refer to Loop Closures