XGRIDS Pro Guide™ / Module 3: Field Technique

3.4 Challenging Environments

Standard technique works in standard environments. Long corridors, large open spaces, glass, darkness, and outdoor conditions each put specific pressure on SLAM that requires a deliberate adjustment to how you scan. This page covers each environment type with the specific technique that keeps the scan recoverable.

Section 1

Long Corridors and Tunnels

Long, straight corridors and tunnels are one of the most demanding SLAM environments because they are geometrically repetitive and visually degenerate. Every cross-section of the tunnel looks nearly identical to the previous one, which makes feature matching unreliable over distance. Both LiDAR and visual tracking degrade progressively in these conditions.

The practical limit for a single continuous scan run in a narrow corridor or tunnel is 1,600 feet. Beyond that, the combination of LiDAR degeneracy (repeating geometry) and visual degeneracy (repeating textures, low light) makes drift accumulation too great to correct reliably in post-processing. A 1 km tunnel tested by XGRIDS required four runs of 173 m, 218 m, 250 m, and 378 m respectively, each under 500 m, and was then merged into a single coordinate system using Map Fusion. The resulting point cloud showed no distortion, divergence, or drift at the segment joins.

Enable Narrow Scene mode in LixelStudio before processing any tunnel or long corridor scan. This SLAM mode is specifically tuned for mine shafts, tunnels, and long indoor corridors. It increases reconstruction success within 1,600-foot segments. Do not use it for regular indoor environments — it can cause failure in standard scenarios.

Tunnel and Long Corridor Procedure

Plan Segments of No More Than 1,600 Feet Each

Divide the total corridor or tunnel length into segments before scanning. Each segment must be processed as a separate scan and merged afterward via Map Fusion. Attempting to scan more than 1,600 feet in a single continuous run in a degenerate environment is not recoverable if tracking fails mid-run.

Place Control Points at Segment Boundaries

For the first segment: place 2 control points at the far end only (none at the start). For the last segment: place 2 control points at the near end only. For all middle segments: place 2 control points at both the start and the end, 4 total per middle segment. Minimum spacing between any two points is 15 feet.

Match Point IDs and Positions Precisely Across Segments

Any physical location used as a shared control point between two segments must have the same point ID in both scans, typed identically. When marking the shared point in the second scan, the device position and orientation must be within 4 inches and 10 degrees of where and how it was held in the first scan. This consistency is critical to Map Fusion succeeding in degenerate environments.

Walk Slowly and Keep the Device Vertical

In low-light tunnels, slow to 1.5 ft/s or below. The visual cameras are working harder than normal to find trackable features against repeated texture and low illumination. Reduce your speed further if you see the tracking indicator show stress in LixelGO.

Process Each Segment With Narrow Scene Mode, Then Merge

Process each segment individually in LixelStudio with Narrow Scene SLAM mode selected. After all segments are processed successfully, use Map Fusion to merge them using the shared control points at segment boundaries.

Office Corridors and Hallways Under 500 Meters

Standard office corridors that are long but not geometrically degenerate do not require segmentation. The key technique is serpentine movement rather than walking straight down the center. A serpentine path weaves gently from wall to wall, capturing geometry from multiple angles and creating coverage overlap that supports more reliable tracking than a single centerline pass.

For any corridor over roughly 160 feet (SLAM-only), plan at least one deliberate loop by branching into a room and returning. Do not walk the full length in one direction without a loop closure opportunity. With RTK Fixed status, corridors up to 500 feet are viable because coordinate drift is bounded by the continuous position stream.

Section 2

Large Open and Featureless Spaces

Warehouses, exhibition halls, parking structures, airport concourses, and large industrial floors present SLAM with the opposite problem from tunnels: instead of too much repetitive geometry close up, there is too little geometry at close range. The scanner can see distant walls and the ceiling, but features per unit distance are sparse. SLAM drift accumulates faster in open spaces because loop closure opportunities are fewer and feature matching is less dense.

Serpentine Passes

Walk S-shaped routes across the space rather than parallel straight lines. Each curve in the serpentine creates a direction change that generates feature matching from a new angle, and adjacent passes overlap to reinforce coverage. Straight parallel passes provide no loop closure and minimal cross-referencing between passes.

Temporary Reference Markers

Place temporary physical markers, surveying targets, cones, or distinctive objects, at regular intervals across the space before scanning. These give SLAM identifiable features to match against as you move through the otherwise featureless area. Control point stickers double as both SLAM targets and georeferencing anchors.

Frequent Control Points

In large open spaces, plan ground control points closer together than the standard maximum spacing. The standard rule is 330 feet or less for the L2 Pro and 165 feet or less for the K1. In featureless environments, use the tighter end of these ranges to give LixelStudio more correction anchors to work with during processing.

Slower Speed Throughout

The standard 3 ft/s limit applies to environments with adequate features. In genuinely featureless open spaces, slow to 1.5 ft/s even where there are no other speed constraints. More scan data per meter gives SLAM more matching material to work with against sparse features.

Section 3

Reflective and Glass Surfaces

Mirrors, glass walls, polished concrete floors, stainless steel panels, and any specular surface introduce geometry that does not physically exist: the reflections appear as real structure to the cameras and, to a lesser extent, to LiDAR. Repeated passes in front of reflective surfaces compound the problem by building up multiple conflicting representations of the same reflected geometry.

Capture reflective surfaces in a single sweep, not multiple passes. One deliberate pass at the correct angle introduces a manageable amount of reflection artifact. Three or four passes of the same mirror accumulates artifacts that significantly degrade the point cloud in that area
Approach at an angle. A direct perpendicular pass in front of a mirror or glass wall returns the full reflection of yourself and the device directly back at the cameras. An angled approach reduces the direct reflection exposure
Keep moving at 1.5 ft/s past any reflective surface. Slowing down or pausing in front of a mirror or glass panel increases dwell time in the reflection zone, exactly where you want to minimize it
Low-texture surfaces (plain white walls, frosted glass) are not a reflection problem but a tracking problem. Minimize time spent scanning predominantly in front of large uniform surfaces where the visual camera has nothing to track. Move through these zones continuously
In high-reflection spaces (lobbies with mirror walls, server rooms with polished floors), plan your route to minimize the proportion of the total scan time spent adjacent to the reflective surfaces

Section 4

Dark and Low-Light Areas

LiDAR is not affected by darkness, it is an active sensor that generates its own light. The visual cameras in XGRIDS devices are passive sensors that depend entirely on ambient illumination. In dark conditions, LiDAR tracking continues normally, but visual tracking degrades or fails. Since Multi-SLAM fuses all three sensors, visual degradation weakens the overall tracking quality even though LiDAR is unaffected.

Turn on all available lights before scanning any interior space. Open curtains and blinds. The scanning environment should be as evenly lit as possible. Uniform lighting is better than bright spots, avoid situations where one area is brightly lit and adjacent areas are dark
Use supplemental portable lighting in spaces where fixed lighting is inadequate. A second person carrying a portable LED panel and walking ahead of the operator is effective. The panel should illuminate the area the operator is about to enter, not the area they are currently scanning
Slow to 1.5 ft/s or below in any dark area. The visual camera's effective frame rate for tracking is reduced in low light — more scan data per meter partially compensates for lower visual tracking confidence
Avoid abrupt lighting transitions. Walking from a well-lit area directly into darkness gives the camera system a hard cut with no time to adapt. If supplemental lighting is not available, walk the transition slowly to allow the cameras to adjust frame by frame
For spaces that are truly unlit and cannot be illuminated, the point cloud geometry from LiDAR will be captured, but the color data from camera coloring will be missing or very poor quality. Set client expectations accordingly before scanning

Section 5

Outdoor Environments

Outdoor scanning introduces lighting conditions, vegetation, and large open areas that each require specific adjustments. The LiDAR sensor works well outdoors, the L2 Pro has a 300 m range, but camera coloring is highly sensitive to lighting quality, and SLAM tracking in open outdoor environments faces the same feature-sparsity challenges as large interior open spaces.

Scan outdoors on overcast days when possible. Overcast conditions provide diffuse, even illumination that the camera system handles well. Direct sunlight creates harsh contrast, shadows with no detail, and overexposed highlights that all degrade camera colorization. The best time on a sunny day is around solar noon when shadows are shortest, but an overcast day is preferable to any sunny condition
Never point the scanner toward the sun. Orienting the device so cameras face the sun causes severe overexposure in the captured images. Position yourself so the sun is behind or to the side at all times
Vegetation is not accurately represented in point clouds. Trees, shrubs, and tall grass produce sparse, noisy geometry because LiDAR pulses penetrate partially through foliage and return from multiple depths. Do not use vegetation geometry for measurement. For projects where vegetation needs to be documented, note this limitation to clients before scanning
Use RTK for outdoor projects where absolute accuracy is required. Outdoor environments often lack the interior wall and ceiling geometry that helps SLAM maintain tracking. RTK provides an external coordinate anchor that reduces the impact of drift in open areas
Apply serpentine routes and plan loop closures deliberately, outdoor areas have fewer natural loop closure opportunities than buildings. Plan explicit return passes through previously-scanned areas rather than assuming coverage will create closures
Set the Point Cloud Participation Rate (PPR) to Low in LixelStudio for outdoor scans. The default Normal setting can cause sky color to bleed into tree edges and building tops in outdoor point clouds. Low PPR reduces this artifact

Section 6

Dynamic Scenes and Pedestrian Management

People, vehicles, and moving equipment introduce geometry that changes frame to frame. SLAM assumes the environment is static, objects that move between frames create conflicting feature matches that degrade tracking. LixelStudio's dynamic object removal handles most single-frame people and vehicles, but the algorithm has limits.

Scan occupied spaces with as few people as possible. Clear the scan path before beginning a session in active workspaces. Coordinate with facilities staff to keep corridors and rooms clear during the scan window
Do not allow anyone to walk alongside the operator at matching speed. A person who keeps pace with the operator appears to the dynamic object removal algorithm as a static object, their relative velocity to the scanner is near zero. They will be retained in the point cloud as permanent geometry
People walking through the scan path at perpendicular or oblique angles are generally handled correctly by the dynamic object removal algorithm. The problem is specifically people moving at the same speed and direction as the scan
For environments where people cannot be cleared (active retail, transit hubs, construction sites with workers), slow down and plan multiple overlapping passes through populated areas. Processing will use the multi-pass data to identify and remove dynamic objects more effectively than a single pass through a crowded area
Dynamic screens, construction equipment in operation, and traffic all introduce large-scale, rapid changes to the camera frame. Avoid scanning with these in the foreground, route around active equipment zones and plan scans of exterior areas with traffic during off-peak periods

One more environment with its own specific demands before leaving this module: data centers, where repetitive geometry and frequent rescans require a purpose-built approach.

Data Centers →