LiDAR-based localization environment requirements
Assessment principles mid360 LiDAR based localization requires not only "objects in the surroundings," but also sufficient and stable geometric features in the environment. Stru…
7 min read · English documentationAssessment principles
mid360 LiDAR-based localization requires not only "objects in the surroundings," but also sufficient and stable geometric features in the environment. Structures such as walls, columns, wall corners, beams, door frames, and fixed equipment constrain the UAV's position and attitude from different directions. Such an environment is described as having sufficient geometric constraints and good observability. When only the ground or a single wall is present, the space is too open, or structures repeat over long distances, geometric degeneracy (LiDAR Degeneracy) can easily occur, causing localization to drift or diverge.
Caution
1. All subsequent mid360 LiDAR-based localization functions must meet the environmental requirements in this chapter. Do not arm for takeoff if the positioning data is abnormal or continues to drift, or if the point cloud is visibly misaligned.
2. Conditions such as high temperatures, low temperatures, severe vibration, and dense fog may reduce the detection performance of the mid360; rain, fog, and dust may also produce interfering echoes.
3. Detection performance may not be guaranteed for low-reflectivity objects 0.1–1 m from the LiDAR or for small objects.
For more information about detection ranges, target reflectivity, and environmental limitations, see the official Livox MID-360 specifications.
1. How to determine whether an environment is suitable
| Assessment item | Technical explanation | On-site assessment method |
|---|---|---|
| Are there a variety of fixed structures? | Geometric feature richness | A variety of structures, such as walls, columns, wall corners, beams, door frames, shelves, or fixed equipment, should be visible around the UAV, rather than only the ground. |
| Do the structures come from multiple directions? | Multidirectional geometric constraints and observability | Fixed structures should be present in the front, rear, left, and right directions; there should be at least walls or columns facing different directions. A single plane does not provide sufficient constraints. |
| Do the structures vary in height? | 3D geometric constraints | The environment should contain door frames, beams, equipment at different heights, or other 3D structures; the structures must not all be concentrated at the same height. |
| Are point-cloud echoes stable? | Valid echoes and point-cloud continuity | In the ground control station or RViz, the outlines of walls and columns should remain continuous and stable, without large missing areas, flickering, or ghosting. |
| Is the environment relatively static? | Dynamic interference | Most points in the point cloud should come from fixed objects. Large numbers of pedestrians, moving vehicles, or vegetation swaying in the wind reduce the reliability of point-cloud matching. |
Rule of thumb: If there is only ground, sky, or one long wall around the UAV, the environment is generally unsuitable for LiDAR-based localization. If fixed walls, wall corners, columns, and beams are visible in multiple directions, the environment is generally more suitable for LiDAR-based localization.
2. Common suitable environments
| Common environment | Why it is suitable | Conditions for use |
|---|---|---|
| Underground parking garage | Walls, columns, beams, wall corners, and parked vehicles can provide relatively rich planar features, edge features, and multidirectional geometric constraints. | Give priority to areas with little vehicle and pedestrian traffic and normal lighting and ventilation. Do not fly directly in large open floors or long lanes with completely repetitive structures. |
| Office, laboratory, or large indoor room | Wall corners, door frames, furniture, and fixed equipment can provide nearby geometric constraints. | The flight space must meet safety-distance requirements. If there are many glass partitions, check the point-cloud quality first. |
| Near buildings, in courtyards, or in semi-open spaces | Building facades, wall corners, columns, tree trunks, and fixed facilities can compensate for insufficient features in open areas. | Do not move away from fixed structures into a large open area. If leaves are swaying noticeably, vegetation must not be used as the primary localization feature. |
| Tunnels or mines with noticeable structural variation | Irregular rock walls, support structures, equipment, junctions, and cross-section changes can provide effective constraints. | Use only when the point cloud is stable and the structures vary significantly. Long, straight tunnels with a uniform cross-section are high-risk degenerate environments. |
Using an underground parking garage
Underground parking garages are generally suitable environments for mid360 localization, but being in an "underground parking garage" does not by itself guarantee safe localization. Select an area that has walls, columns, beams, wall corners, and stationary vehicles so that the LiDAR can obtain stable echoes in multiple directions. Use is still not recommended in the following garage areas:
- Large open floors in newly built garages where vehicles have not yet been parked;
- Long, straight lanes with completely identical structures on both sides;
- Areas near entrances or exits with intense light, rain, fog, or significant airborne dust;
- Highly dynamic areas with continuous vehicle and pedestrian movement;
- Areas containing large surfaces made of glass, specular metal, or black light-absorbing materials.
3. Unsuitable or high-risk environments
| Environment type | Cause of risk | Main risk |
|---|---|---|
| Open playgrounds, fields, rooftops, and large empty warehouses | Feature-poor environments and insufficient geometric constraints | The LiDAR can primarily see only the ground, so horizontal constraints are insufficient and localization is prone to drift. |
| Long, straight corridors; long, straight tunnels; and aisles with completely repetitive shelving | Geometric degeneracy and repetitive-structure degeneracy | Displacement along the aisle or tunnel is difficult to estimate reliably, and repetitive structures may also cause incorrect matching. |
| Large surfaces made of glass, mirrors, polished metal, or water | Specular reflections, multipath reflections, and invalid echoes | Points in the point cloud may be missing, jump abruptly, or appear in incorrect positions. |
| Black light-absorbing materials, dark fabric, thin wires, and mesh | Low reflectivity and sparse echoes | Too few points are returned, and small obstacles may also not be detected consistently. |
| Crowded areas and areas with continuously moving vehicles or frequently moving robotic arms | Highly dynamic scenes | Dynamic point clouds disrupt matching between adjacent frames and may cause localization jitter or drift. |
| Dense fog, heavy rain, airborne dust, or smoke | Atmospheric particle interference and backscatter | Airborne particles generate interference points and reduce the detection quality of valid targets. |
| Severe vibration or an insecurely mounted LiDAR | Extrinsic-parameter changes and motion distortion | The relative position between the LiDAR and the airframe changes, which may distort the localization results. |
4. Reflectivity and object materials
| Recommended surfaces | High-risk surfaces |
|---|---|
| Objects with stable diffuse reflection, such as concrete walls, brick walls, wooden furniture, cardboard boxes, and ordinary painted surfaces | Black light-absorbing materials, fabric, thin wires, mesh, glass, mirrors, polished metal, and water surfaces |
Do not rely solely on a single material or plane as a localization reference. Even if the objects have good reflectivity, geometric degeneracy may still occur if all structures face the same direction or the space is too open.
5. Preflight on-site inspection
- Place the UAV at the planned takeoff point and start mid360 LiDAR-based localization, but do not arm it yet.
- View the point cloud in the ground control station or RViz. Confirm that the outlines of walls, columns, and the ground are continuous and stable, without large missing areas or noticeable ghosting.
- Keep the UAV stationary and observe the position data. Confirm that x, y, and z do not continuously drift in one direction or jump suddenly.
- Slowly move or rotate the UAV for inspection. Confirm that the point-cloud map does not show layering, tearing, or overall misalignment.
- Clear all personnel and moving vehicles from the flight area, and confirm that fixed structures are always present around the flight path.
- When testing in a new environment for the first time, begin with a low-altitude, small-area hover test. If localization is abnormal, stop the test immediately and do not continue with autonomous flight.
6. Temporary methods for improving the environment
| Problem | Temporary improvement method |
|---|---|
| The space is too open or lacks features | Place 3D objects such as cardboard boxes, columns, and fixed markers with distinct edges in different directions and at different distances. Avoid arranging all markers at equal intervals or in the same direction. |
| Wall or passage structures are repetitive | Stagger fixed markers with different shapes, sizes, or positions along both sides of the passage to increase structural variation. |
| Surface reflectivity is too low | Apply matte, light-colored stickers to the surfaces of fixed objects or install diffuse-reflective marker boards to increase the number of valid echoes. |
| Glass or mirrored areas cannot be avoided | Temporarily cover them with opaque, matte materials and add fixed markers with distinct edges nearby. |
After temporarily improving the environment, repeat the preflight on-site inspection. Adding markers can only improve LiDAR observability; it cannot replace safety distances, flight protection, or human monitoring.
