2.5 Critical Basics

SLAM Drift

SLAM tracks position by matching what it currently sees to what it has previously seen. Every match introduces a small amount of error. Over distance, those small errors compound. The scanner does not know they are accumulating; it has no external reference to compare against. This compounding positional error is called SLAM drift, and it affects all mobile LiDAR systems regardless of manufacturer or price.

XGRIDS Multi-SLAM fuses three sensors simultaneously, LiDAR, visual cameras, and IMU, which reduces the rate at which drift accumulates compared to single-sensor systems. It does not eliminate drift. The IMU alone introduces a leveling error of approximately 0.07 degrees. SLAM drift compounds across all three axes (x, y, and z). For most indoor projects, the effect is modest. For large-scale or long-duration scans without control, it becomes significant.

Loop Closure

Loop closure is the primary mechanism for correcting accumulated drift. When the scanner returns to an area it has previously scanned, the SLAM algorithm compares the current geometry to the stored geometry of that location, measures the discrepancy, and distributes a correction back across the stored trajectory. The entire path between the two visits is adjusted.

This is why a scan route that creates genuine loops produces more accurate data than a straight-line route of the same length.

Loop closure has two requirements. Both must be met.

1. You physically return to an area you have already scanned.
2. Your viewing angle when you return is within 40 degrees of your original angle at that location. Returning from the opposite end of a corridor often fails this condition; the geometry looks different enough from the reversed direction that the match fails.

A route that ends at the starting point after traveling the entire perimeter from only one direction does not reliably trigger loop closure. Plan routes with multiple crossing loops, not a single perimeter circuit.

When to Create Loops

• At any point where you can return through an area you previously walked, do it deliberately, not incidentally
• At the end of every scan session, return near the initialization point from a similar angle to your starting position
• During large-scale scans, plan at least one midway loop before you have traveled more than a few hundred meters (roughly a thousand feet) from your start
• In challenging environments (long corridors, large open areas), loop more frequently, not less

Relative vs. Absolute Accuracy

Conflating the two leads to incorrect deliverables and client expectations that cannot be met.

Relative Accuracy

Internal consistency of the point cloud. Whether a wall is planar. Whether measured dimensions repeat between two measurement points. Whether the geometry holds across the scan without distortion. Achieved through good technique and loop closure. Does not require GPS, RTK, or GCPs. The published relative accuracy figures (≤1 cm (0.4 in) RMSE for the L2 Pro, ≤1 cm (0.4 in) for the K2) are achievable with technique alone. For the L2 Pro, the figure applies between points within 100 m (328 ft); for the K2, within 10 m (33 ft).

Absolute Accuracy

How closely the point cloud registers to a real-world coordinate system. Whether a point in the scan corresponds to a verifiable coordinate on earth. Relative accuracy alone cannot produce this; it requires RTK, PPK, or surveyed ground control points applied during or after the scan. The published absolute accuracy figures (≤3 cm (1.2 in) RMSE for the L2 Pro; 3 cm (1.2 in), max 5 cm (2 in), for the K2) require RTK or GCPs. For the L2 Pro, keep continuous RTK gaps under 100 m (328 ft). On the K2, keep continuous RTK gaps 164 ft (under 50 m).

What the Specifications Mean, and What They Do Not

The accuracy figures in XGRIDS documentation are achievable results under proper conditions, not guaranteed outcomes from any scan. A scan with poor technique, no loop closure, and no control can produce data far outside these figures.

Relative accuracy applies between points within 100 m (328 ft) for the L2 Pro and within 10 m (33 ft) for the K2. Absolute accuracy requires RTK or GCP coverage; keep continuous unfixed gaps to 328 ft (under 100 m) on the L2 Pro and 164 ft (under 50 m) on the K2. The L3 figures come from a different test: its relative accuracy is stated at 1 sigma for a post-processed capture in a 5,382 sq ft (500 m²) indoor test environment, not as RMSE, and its absolute accuracy is RMSE under open sky.

Repeat Accuracy and Drift at Distance

Repeat accuracy measures session-to-session consistency: if you scan the same space on two different days, how closely do the results align? This matters for progress monitoring and change detection workflows. The published repeat accuracy figures assume both sessions used RTK with no disconnection. SLAM drift at distance indicates how much positional error accumulates over a given path length without loop closure correction.

The L2 Pro SLAM drift figure applies at 100 m (328 ft). XGRIDS does not publish a K2 drift-at-distance figure. For control point spacing, the L2 Pro maximum is 328 ft (within 100 m of each other) and the K2 maximum is 164 ft (within 50 m of each other).

Decisions You Cannot Undo After the Scan

Most scanning mistakes are recoverable. Some are not. These are the decisions that must be made correctly in the field; they cannot be corrected in LixelStudio or LCC Studio after the fact.

• Georeferencing strategy. RTK must be configured and active before scanning begins. GCPs must be physically placed in the environment and marked in LixelGO during the scan. There is no way to add georeferencing to a completed scan after the session ends. If a client requires absolute accuracy and you scanned without RTK or GCPs, the data cannot be corrected; the project must be rescanned.
• Initialization quality. If the initialization position was poor or the device was moved during the 15-second countdown, the starting reference frame is corrupted. The error propagates through the entire session. This cannot be corrected in processing. The scan must be redone from the beginning.
• Coverage gaps. Areas that were not scanned do not exist in the data. Processing cannot fill in geometry that was never captured. If a room was missed or a floor was partially skipped, that area requires a return visit. The LixelGO preview in the field is the only practical opportunity to catch this before leaving site.
• GCP naming. The names you type in LixelGO when marking ground control points during the scan must exactly match the point names in your coordinate CSV file, character for character, including capitalization. A mismatch stops LixelStudio from applying the coordinate to the point. It can be corrected after the scan by supplying a new spreadsheet of control point names, so it does not cost a return visit.
• Scan duration relative to RAM. If a single scan exceeds what your processing machine's RAM can handle, the processing software will fail mid-job. In LCC Studio, 64 GB reliably reconstructs 30 minutes of capture data and 128 GB reconstructs 60 minutes of capture data. The scan data is not lost when this happens, but you cannot process it as a single session. You must split it and use Map Fusion, which requires additional planning and processing time that was not in the original project schedule. Additionally, do not run LCC Studio and LixelStudio processing jobs simultaneously on the same machine. Process one pipeline to completion before starting the other.

Before You Leave Site

• The scan stopped cleanly; the LED reached solid green before the device was powered off
• The project appears in LixelGO with a reasonable duration and file size for the area covered
• The point cloud preview shows no large coverage gaps in areas you walked through
• If RTK was used, the session included more than 100 valid RTK-fixed points and more than 10 m (33 ft) of movement while in Fixed status. LixelStudio requires both conditions to apply the coordinate transformation
• If GCPs were used, the names logged during the scan match your coordinate file exactly, including capitalization
• Battery state is noted; if a battery warning fired during the session, check the scan duration to confirm the data captures the full intended area

Module 3 covers field technique in depth: route planning, movement fundamentals, challenging environments, and coverage strategy for every project type.

Movement and Posture →