3.4 Challenging Environments

Tunnels and Long Corridors

Tunnels combine three SLAM challenges in one environment: repetitive geometry, low feature density, and length that exceeds the comfortable range of a single uncorrected session. They require segmentation, target-based alignment, and a specific processing mode that defaults off.

Segment Length

• Maximum 1,600 ft (500 m) per scan segment. Divide longer tunnels into segments before starting. A continuous tunnel scan beyond this length is not reliably recoverable in post-processing
• Plan segment boundaries at structural features where possible: cross-passages, recesses, equipment alcoves. These give the geometry that distinguishes one segment's start from another segment's end

Speed and Posture

• 1.6 ft/s (0.5 m/s) or below throughout. In low-light tunnels, slower is better
• Side shuffle in narrow tunnels under 8 ft (2.5 m) wide. This keeps both walls in the LiDAR field of view at all times
• Forward walk in wider tunnels at the same speed limit. The wider geometry gives SLAM more lateral feature variation

Multi-Segment Overlap

Where a tunnel must be split into multiple sessions, each pair of adjacent segments must share an overlap zone with marked shared points. Map Fusion uses these to align the segments into a continuous tunnel.

• Plan a shared overlap point at every segment boundary. The same physical location appears at the end of one segment and the start of the next
• When marking the shared point in the second scan, device position must be within 4 in (10 cm) and 10 degrees of its position in the first scan. Mark the spot on the ground with tape during the first segment so you can return to it precisely for the second
• Minimum spacing between shared points: 15 ft (5 m). Closer placements provide overlapping reference data that confuses alignment rather than improves it

Control Point Placement Across Segments

Multi-segment tunnels with georeferencing need control points distributed across the tunnel length, not just at the entrance. The pattern below ensures every segment has georeferencing data while minimizing field placement time.

• First segment: two control points at the end only
• Last segment: two control points at the start only
• All middle segments: two control points at the start and two at the end (four total per segment)
• Give the same physical location the same point ID in every segment where it appears. This is how Map Fusion identifies the same point across multiple sessions

Target Placement

Tunnels rarely have enough natural geometric variation for SLAM to track confidently across long distances. Place high-contrast SLAM anchor targets every 100 to 150 ft (30 to 45 m) in long, featureless runs.

• The L2 Pro ships with magnetic steel SLAM anchor targets that attach to ferromagnetic surfaces (rebar, embedded steel, metal panels)
• The K2 ships with 30 reflective sticker targets that adhere to any flat surface
• These targets serve as SLAM anchors, not GCPs. They give the system high-contrast features to match against in environments where the surrounding geometry is uniform. If you also need absolute coordinates, supply those targets with surveyed coordinates in LixelStudio, at which point they become GCPs

Processing

Reflective Surfaces: Glass, Mirrors, Polished Metal

Reflective surfaces are the most common cause of indoor SLAM artifacts. The LiDAR's pulses pass through glass and bounce off mirrors, creating the appearance of geometry behind or beyond surfaces that physically end at the glass. The visual cameras can track reflections as false features.

Distance and Speed

• Maintain at least 3.3 ft (1 m) distance from highly reflective objects where the scan path allows. Closer proximity increases the LiDAR's reflection exposure
• Move at 1.6 ft/s (0.5 m/s) past any reflective surface. Slower past large glass facades; faster past small mirrors is acceptable but not preferred
• Never stop in front of glass or mirrors. Standing still in front of a reflective surface accumulates many observations of the false geometry that the reflection produces. A continuous walk past produces fewer false observations and lets SLAM identify them as outliers

Approach Angle

• Approach glass walls at an angle rather than head-on. A 30 to 60 degree angle of approach reduces direct LiDAR reflection back into the sensor. Head-on approach maximizes reflection
• Capture mirror and glass surfaces in a single sweep. Multiple passes in front of the same reflective surface introduce conflicting reflection observations that degrade the surrounding point cloud
• For fully-glazed office facades, plan your route to cross glass zones rather than walk parallel to them at close range. Walking parallel to glass at 6 ft (2 m) for an extended distance is the worst case; crossing it perpendicular at the same distance is significantly better

Body Positioning

Where you cannot avoid proximity to a mirror, position your body between the scanner and the reflective surface. Your body absorbs LiDAR pulses that would otherwise create false reflection data. This is particularly useful in residential bathrooms with vanity mirrors and small rooms with mirrored closet doors.

PortalCam Sensitivity

The PortalCam is more severely affected by reflective surfaces than the L2 Pro and K2 because its 3DGS reconstruction depends more heavily on visual data. Reflections that produce minor LiDAR artifacts can produce significant visual artifacts in the 3DGS output. In environments with extensive reflective surfaces, consider scanning with an L2 Pro or K2 for point cloud deliverables and supplementing with PortalCam coverage of less reflective areas if 3DGS output is required.

Dark Environments

Dark environments degrade two of the three SLAM inputs simultaneously: the visual cameras have insufficient light for feature tracking, and the LiDAR has less ambient reflectivity to bounce off. The IMU continues working unchanged, but with two of three inputs degraded, drift accumulates faster.

Pace and Lighting

• Slow to 1.6 ft/s (0.5 m/s) or below in any dark area. Slower is better in low-light corridors and rooms
• Bring a portable light panel. A handheld LED panel held by the operator or an assistant significantly improves both visual tracking and 3DGS colorization quality. Even 200 to 400 lumens of well-aimed fill light makes a measurable difference
• Angle the light forward into the area you are walking toward, not downward at the floor at your feet. SLAM is continuously matching new geometry against what it has already seen. The upcoming area needs to be illuminated before you reach it, not after
• For static dark spaces (utility rooms, basements, electrical closets), consider setting up battery-powered work lights before scanning. Stationary lights produce more even illumination than handheld panels and free the operator to focus on posture

Anchor Targets in Dark Areas

• Place high-contrast SLAM anchor targets in dark featureless areas before scanning. L2 Pro magnetic steel targets or K2 reflective sticker targets both serve this role
• Even without surveyed coordinates, these targets give SLAM high-contrast features to match against when visual texture is minimal
• Place targets at intersections, doorways, and along long featureless walls. The goal is to provide reliable matches every 30 to 50 ft (10 to 15 m) of travel

3DGS in Dark Areas

Camera coloring in dark environments produces poor results regardless of operator technique. The cameras need light to capture useful texture data, and aggressive auto-exposure introduces noise and color shifts that 3DGS reconstruction cannot fully compensate for.

• Consider disabling coloring for dark spaces if the deliverable is a point cloud or LiDAR-only output. Re-enable for lit areas
• For 3DGS deliverables in dark spaces, bring sufficient lighting to produce well-exposed visual data. Otherwise, plan dark zones as separate no-color sessions, or accept that dark portions of the model will be visually poor
• Plan dark zones as separate sessions where lighting cannot be improved. Joining a poor-lighting session to a good-lighting session with Map Fusion produces a model with clearly differentiated quality zones, which is often acceptable for deliverables; mixing the conditions within a single session produces a model that looks uniformly compromised

Outdoor Scanning

Outdoor scanning introduces challenges that indoor scanning does not: variable weather, direct sunlight, large open areas with sparse vertical features, and wind. Each affects the scan differently and requires specific technique.

Feature Density

• Outdoor areas require consistent feature density along the scan path. SLAM needs vertical geometry to track against. An open plaza with no surrounding buildings is one of the most difficult outdoor environments
• Include surrounding buildings, walls, fences, or fixed structures in the scan path when crossing open areas. Do not walk a parking lot diagonally with no structures in view; walk along its edges where buildings are visible
• For very large outdoor sites, consider drone-based aerial capture for the open areas and ground-based scanning for the perimeter and entrances. See Module 5: Aerial-Ground Fusion

Sun and Direct Light

• Do not point cameras at the sun or scan into direct sunlight. Overexposure corrupts camera coloring data for the entire scan, not just the overexposed frames
• Plan scan routes to keep the sun at your back where possible. This puts the sun behind the cameras rather than in front of them
• Avoid scanning toward windows in bright sunlight from inside a building. The window appears as an extreme bright source against a darker interior, which overexposes the camera
• Time outdoor scans for diffuse-light conditions where possible: overcast days, early morning, or late afternoon. Direct midday sun produces harsh shadows and overexposure

Wind and Environmental Conditions

• Wind affects PortalCam stability and visual tracking more than the heavier L2 Pro and K2. In windy conditions, walk with the wind rather than against it where possible
• Rain and snow degrade LiDAR performance. Water droplets in the air return false LiDAR pulses, creating noise in the point cloud. Heavy precipitation makes outdoor scanning impractical
• Cold temperatures reduce battery life. Plan shorter sessions or carry additional batteries in cold weather. Keep batteries warm in interior pockets until immediately before use

L2 Pro Range Selection for Outdoor Work

The L2 Pro is available in three range configurations. The 32-300 model has a 985 ft (300 m) maximum LiDAR range and is required for tall structures and large outdoor sites. The 16-120 and 32-120 models are limited to 400 ft (120 m) range, which is sufficient for most indoor and small outdoor work but not for tall facades or expansive exteriors. The K2 has a 100-meter (330 ft) LiDAR maximum range, making it appropriate for indoor and close-quarters outdoor work but not for tall facades or large open exteriors.

Featureless Corridors and Repetitive Geometry

Corridors with uniform walls, identical doorways at regular intervals, and no distinguishing features along their length are SLAM's weakest scenario. The system has nothing to match against. Every position along the corridor looks similar to every other position from the LiDAR's perspective.

Movement Technique

• Side shuffle (crab-walk) in narrow uniform corridors. Keeping both walls in the LiDAR field of view simultaneously is especially important in featureless corridors where SLAM has fewer anchors
• Use diagonal paths in wider featureless spaces rather than walking straight down the centerline. Diagonal movement provides perspective variation that a straight walk does not
• Branch into any available room or recess when traversing long featureless corridors. Even a brief detour into a janitor's closet creates a loop closure that resets accumulated drift

Target Strategy

• Place anchor targets at irregular intervals in long featureless corridors. Regular spacing matches the existing repetitive geometry; irregular spacing gives SLAM distinguishable anchor points
• Use different target placements between similar corridors. If a building has multiple identical corridors on different floors, vary the target height or wall position on each floor so the floors are distinguishable to SLAM

When Multiple Challenges Combine

A dark featureless corridor with reflective windows on one side combines three of the challenges on this page. A glass-walled outdoor walkway with sun glare combines reflective surfaces with outdoor lighting. These compound environments require techniques from each contributing category.

The general rule is: apply the most conservative technique from each applicable category. Slow to the slowest applicable speed, use anchor targets even if only one of the contributing categories normally requires them, and budget more time. A compound-challenge environment that takes 10 minutes to scan in normal conditions may take 25 minutes with proper technique. The alternative is data that does not process cleanly.

Plan compound-challenge zones explicitly. They warrant a separate line item on the scan plan, with their own segment boundaries, target placements, and time allocation. Surprises in the field at compound-challenge zones are how unrecoverable scans happen.