This article explains what drift is, why it occurs during SLAM-based scanning, and how to minimize its impact through best-practice scanning techniques.
What is drift
Drift is the gradual stretching or warping of a point cloud that can occur as SLAM incrementally estimates its position while mapping an environment.
Why does drift happen?
Hovermap uses LiDAR to determine its position within its environment. It emits laser pulses and then measures how long it takes for the light to return. Based on the time this takes, Hovermap knows how far away an object is from the sensor. Drift occurs as a result of tolerances in measuring these light return times. The longer a scan takes, and the further the scan travels, the more drift is accumulated.
Some environments, such as those with a lot of vegetation, or long scans that only go in one direction (such as a tunnel or a road), can lead to more drift accumulating in your final scan. Refer to the Common scenarios section for more specific information on best practices for scanning in these situations.
Why it matters
Drift directly affects the accuracy and reliability of scan data. Excessive drift can distort geometry, reduce alignment accuracy, and impact downstream workflows such as measurement, registration, and integration with other datasets. Reducing drift improves spatial consistency, preserves geometric fidelity, and ensures scan outputs are suitable for analysis, visualisation, and decision-making.
How it works
During a scan, SLAM continuously estimates the sensor’s position by matching newly captured LiDAR data to previously observed features. Each estimate introduces a small amount of uncertainty. Over time, these uncertainties accumulate, resulting in drift.
Techniques such as revisiting previously scanned areas, introducing changes in scan direction, and limiting scan duration help SLAM correct its position estimates and reduce accumulated error.
Common scenarios
Drift considerations are especially important in environments where scans are long, complex, or feature-poor.
Long, linear scans can accumulate drift due to limited geometric variation. Looping or segmented scans help reduce error.
Irregular or moving features can reduce SLAM stability. Shorter scans with repeated passes improve consistency.
Sparse features can make localisation more difficult. Structured scan patterns help maintain accuracy.
Environments with repeating shapes or uniform structures (such as warehouse aisles or pipe racks) provide limited unique features for SLAM, increasing the likelihood of drift. Introducing varied scan paths and revisiting previous areas helps improve positional correction.
Best Practices to avoid Drift
Applying these best practices helps minimize drift and improve the accuracy and consistency of your scan results.
Do | Don't |
|---|---|
Limit your scans to the smallest area possible. Long scans can cause significant drift. | Conduct unnecessarily long scans that cover large distances without revisiting previous areas. |
“Close the loop” or scan in a grid pattern to average out drift. For more information, refer to the Scan patterns section. | Scan continuously in a single direction without returning to previously scanned areas. |
Break large environments into multiple shorter scans where possible. | Attempt to capture very large environments in a single continuous scan. |
Important Considerations
Drift is cumulative and cannot be fully removed after capture
Environmental conditions significantly affect SLAM performance
Longer scan duration increases the likelihood of drift
Best practices should be applied during capture, not relied on in post-processing