Pushing the Limits of SLAM: 90,000 m² Underground, 2 Levels, Zero Drift — How the MVP S2 Did It in 1 Hour

Peter Yang, Tersus GNSS      06 February, 2026

1. Scan. 0 GCPs. 40x Efficiency. When other algorithms fail in feature-poor environments, the MVP S2 thrives.

In the world of mobile mapping, underground parking lots are often called "SLAM killers."

The combination of extreme low light, long corridors with repetitive features (hundreds of identical pillars), and the total absence of GNSS signals creates a "perfect storm" for algorithm failure. In these environments, most handheld SLAM devices suffer from trajectory drift, loop closure failures, and the dreaded "double wall" effect (layering). To compensate, surveyors are usually forced to deploy dense Ground Control Points (GCPs)—a costly and time-consuming process.

But recently, in a Smart City Digital Twin project in China, the Tersus MVP S2 took on an extreme challenge: Scanning a 90,000 m² (approx. 970,000 sq. ft.) two-level underground parking complex in a single continuous session, without placing a single GCP.

The result? It didn't just survive; it redefined the standard for underground efficiency.

01. The Challenge: A Geometric Nightmare
The project required a high-precision digital twin of the B1 and B2 parking levels of a bustling commercial complex.

A. Scale: 90,000 m² across two floors.

B. Environment: Dim lighting, heavy vehicle traffic, and highly repetitive geometry (columns) that typically confuses SLAM feature matching.

C. Constraint: The facility was in active use. Shutting it down for a week of static scanning was impossible.

The Traditional Bottleneck:

Using a Terrestrial Laser Scanner (TLS) would have required over 500 individual setups. Even for a skilled crew, this would take at least 5 to 7 days of fieldwork, plus several days for registration processing.

The "Standard" SLAM Risk:

Without GCPs to "pin" the data, standard SLAM algorithms often accumulate errors exceeding 50 cm over such long distances. This results in map layering where the B1 and B2 floors don't align, or the start and end points of the loop don't close.

02. The Solution: "Trust Your Algorithm."
The team deployed the Tersus MVP S2 Mobile SLAM Scanner in Backpack Mode to tackle the site.

A. The Workflow: The operator started outdoors to initialize the trajectory with a high-precision RTK fix. Then, they walked down the ramp into the B1 and B2 levels, transitioning instantly into a pure SLAM environment.

B. The "Blind" Scan: For 1 hour, the device operated without any satellite signal or external control points. It relied entirely on its 32-channel LiDAR sensor and the Tersus self-constrained SLAM algorithm to navigate the complex.

C. Hot-Swap Capability: Thanks to the dual-battery hot-swap design, the operator could scan continuously without powering down, maintaining a single, unbroken trajectory file.

03. The Results: Data That Speaks for Itself
After just 1 hour of field work, the raw data were processed in the MVP S2 Mapper software. The results stunned the surveying team.

1. 40x Efficiency Boost

From 1 week (static scanning) to 1 hour (MVP S2). The dramatic reduction in field time meant minimal disruption to the parking lot's operations and significantly lower labor costs.

2. "Zero Drift" Precision

We analyzed the most critical stress points—the spiral ramps connecting floors and the furthest corners of the lot:

A. No Layering: The walls and columns appeared as single, sharp surfaces. The point cloud thickness was consistent at ≈1 cm.

B. Perfect Alignment: The B1 and B2 levels aligned perfectly in the Z-axis, with no vertical drift despite the lack of GNSS.

C. Absolute Accuracy: Even after kilometers of walking without GNSS, checkpoints confirmed the absolute accuracy remained within the 3 cm range, qualifying it as true survey-grade data.

3. Intelligent Noise Removal

The parking lot was active with cars moving in and out. The MVP S2’s dynamic object filtering automatically identified and removed "ghosts" of moving vehicles and pedestrians, leaving behind a clean structural model ready for vectorization.

(Caption: The complete 90,000 m² point cloud. Note the clean, continuous trajectory with no breaks.)

(Caption: Cross-section analysis showing wall thickness of approx. 1 cm with zero layering—proof of a robust algorithm.)

04. Why the MVP S2 Succeeded Where Others Fail
The success of this project boils down to one question: Is your SLAM system relying on RTK to hide its flaws?

Many mobile scanners perform well outdoors but fall apart the moment they lose GNSS. The MVP S2 is engineered with a different philosophy: Robust SLAM First.

A. 32-Channel High-Density LiDAR: By capturing more geometric details (640,000 pts/sec) and having a wider field of view, the S2 finds stable features even in repetitive environments like empty parking garages.

B. Algorithm Self-Constraint: The proprietary algorithm is designed to minimize cumulative error internally, allowing for long-duration "blind" scans without the "banana effect" (warping) common in lesser devices.

Conclusion: Stop Compromising on Underground Projects

For Smart City, Facility Management, and BIM professionals, the MVP S2 proves that you don't need to choose between speed and accuracy.

In a 90,000 m² underground maze, with no control points, no signals, and just one operator, the MVP S2 delivered a flawless digital twin in a single hour. This is the new benchmark for indoor mobile mapping.

About Tersus GNSS Inc.

Tersus GNSS is a leading Global Navigation Satellite System (GNSS) solution provider. Our offerings and services aim to make centimeter-precision positioning affordable for large-scale deployment.

Founded in 2014, we have been pioneers in design and development GNSS RTK products to better cater to the industry’s needs. Our portfolios cover GNSS RTK & PPK OEM boards, David GNSS Receiver, Oscar GNSS Receiver, MatrixRTK [GNSS CORS Systems], and inertial navigation systems.

Designed for ease of use, our solutions support multi-GNSS and provide flexible interfaces for a variety of applications, such as UAVs, surveying, mapping, precision agriculture, lane-level navigation, construction engineering, and deformation monitoring.

Sales inquiry: sales@tersus-gnss.com

Technical support: support@tersus-gnss.com