MVP-SLAM: Multi-Camera Visual-Inertial Floorplan-Prior SLAM
Authors: Asier Bikandi-Noya, Miguel Fernandez-Cortizas, Muhammad Shaheer, Holger Voos, Jose Luis Sanchez-Lopez
Organizations: Authors are with the Automation and Robotics Research Group, Interdisciplinary Centre for Security, Reliability, and Trust (SnT), University of Luxembourg, Luxembourg. · Holger Voos is also associated with the Faculty of Science, Technology, and Medicine, University of Luxembourg, Luxembourg.
Indoor building construction sites are demanding environments for visual SLAM, where variable lighting and repetitive, low-textured structures make the system drift over long trajectories, though structural elements such as walls remain distinguishable despite these conditions. These buildings are constructed according to their as-planned floor plans, available from the design phase, and although the actual as-built site can differ from this design, floor plans still provide a metric reference, both to localize the system in the building and to correct drift. Existing methods often use the floor plan to correct an already-built trajectory offline, and those that instead correct it online typically rely on depth sensors. We instead present MVP-SLAM, an online visual-inertial SLAM on two opposite-facing fisheye cameras that corrects drift from cameras alone by matching walls detected in its map to the floor plan, through a drift-aware policy. A multi-stage integration then turns each matched pair incrementally into a persistent correction, so the trajectory stays corrected and localized within the floor plan as it is built. MVP-SLAM was validated on the multi-floor construction sites of the Hilti-Trimble SLAM Challenge 2026, ranking 2nd of 22 teams in the Localization task (0.29 m mean RMSE) and 5th of 62 teams in the SLAM task (0.24 m), the top-ranked one in both tasks among those that operate online, integrate the floor plan, and localize within it.
Figures & tables
Fig. 1 : MVP-SLAM on a Hilti construction sequence, matching detected walls to the floor plan and correcting the estimated trajectory against ground truth.
Fig. 2 : System architecture of MVP-SLAM. Sensor and floor-plan inputs (left) feed a three-module pipeline whose optimized state is fed back into the SLAM system as persistent plan priors, producing the plan-aligned trajectory estimate (right).
Fig. 3 : Incremental multi-stage integration of a matched wall, preserving earlier corrections.
Floors and Sequences
Ground Floors
Upper Floors
Underground Levels
Floor 1
Floor 2
Floor EG
Floor 3
Floor 4
Floor 5
Floor 6
Floor 7
Floor UG1
Floor UG2
07-07
12-02
12-02
12-03
10-16
12-02a
12-02b
05-19
12-02
05-19
12-02
12-02
06-18
07-07
12-02a
12-02b
12-02a
12-02b
12-03
05-19
06-18
12-02a
12-02b
12-03
12-02
Average
Duration ( sec. )
133.7
276.3
143.7
149.9
242.9
126.4
162.5
115.5
132.7
91.9
193.4
160.5
69.3
73.0
170.7
124.6
117.9
155.1
198.3
205.5
165.8
254.8
219.7
134.4
223.1
161.7
Length ( m )
157.8
321.8
154.8
138.8
240.4
114.9
158.6
128.2
148.8
97.8
214.1
174.9
79.4
72.2
210.6
145.8
145.4
197.1
152.6
262.5
222.6
351.7
322.5
119.1
292.1
185.0
Localization
Map-It Ralph [ 14 ]
71.92
39.68
577.06
233.11
41.86
87.98
320.64
103.89
182.71
38.67
49.77
82.38
49.74
41.31
50.59
109.38
301.10
105.34
69.70
95.13
147.77
264.79
88.25
127.67
–
136.69
TABLE I : Absolute Pose Error (APE) RMSE ( cm ) on Hilti–Trimble SLAM Challenge 2026 sequences, as reported on the official challenge leaderboard on the day of the challenge. Bold and underline denote best and second-best results per column per task; dashes (–) mark unavailable entries for Localization.
Parameter
Symbol
Value
Description
Drift-aware association (Sec. III-C )
score weights
wa,d,f
6 rad -1 , 1 m -1 , 1
angle, dist., overlap
gate ramp
d0,αℓ
1 m, 0.1
offset, drift growth
gate cap
cext/int
5 , 2.5 m
ext./int. distance cap
dist. saturation
dsat
1 m
score distance cap
saturation range
ℓmin/max
15 , 40 m
score ramp
TABLE II : Key parameters of MVP-SLAM.
Fig. 4 : Trajectory comparison in the Localization task, in the floor-plan frame, on three representative sequences.
Reduced
Full
Floor
Sequence
tmetric
Cov.
APE
tmetric
Cov.
APE
Ground
Floor 1 (07-07)
3.1
100
1.38
2.5
100
1.31
Floor EG (10-16)
2.5
100
3.02
2.2
100
2.44
Upper
Floor 3 (05-19)
2.8
100
0.58
2.5
100
0.55
Floor 6 (12-02a)
2.5
100
1.01
2.5
100
0.76
Under
Floor UG1 (06-18)
6.6
96.7
28.0
2.5
100
2.07
TABLE III : Initialization and robustness ablation for Ours (w/o plan) : 3 repeats per sequence, both configurations initialized every run. tmetric : median time to metric scale (s); Cov.: mean coverage (%); APE: median Localization error (m).
Detection
Matching
Floor
Sequence
P
R
F1
P
R
F1
Ground
Floor 1 (07-07)
94.4
50.0
65.4
100.0
94.1
97.0
Floor EG (10-16)
100.0
28.6
44.4
100.0
100.0
100.0
Upper
Floor 3 (05-19)
100.0
57.9
73.3
100.0
100.0
100.0
Floor 6 (12-02a)
84.6
35.5
50.0
90.9
100.0
95.2
Under
Floor UG1 (06-18)
100.0
37.8
54.9
100.0
100.0
100.0
TABLE IV : Wall-detection and wall-matching performance (precision P, recall R, F1, in percent). Overall pools counts.
Monitoring building construction sites requires comparing the as-planned design with the as-built state, which can be estimated in real time using Simultaneous Localization and Mapping (SLAM) techniques. However, visual SLAM is prone to trajectory drift in construction environments, producing maps that are geometrically inaccurate with the actual environment. To address this limitation, we augment an existing RGB-D SLAM system with structural priors derived from the Building Information Model (BIM). The system associates detected walls with their BIM counterparts and includes these correspondences as geometric constraints in the back-end optimization, reducing drift and enhancing global consistency. The proposed method operates in real time and is validated on multiple real construction sites, achieving an average trajectory error reduction of 25.23% and a 7.14% improvement in map accuracy over state-of-the-art baselines. Robustness analyses further demonstrate resilience to incomplete BIM data and geometric discrepancies between as-planned models and the as-built environment.
Asier Bikandi-Noya, Miguel Fernandez-Cortizas, Muhammad Shaheer +3
Automation and Robotics Research Group, Interdisciplinary Centre for Security, Reliability, and Trust (SnT), University of Luxembourg, Luxembourg
Automated progress monitoring on construction sites is an active area of research and development. Robot and human-carried mapping systems have been developed to build 3D maps of building and infrastructure projects. While LiDAR-based mapping systems achieve high accuracy, the cost of LiDAR can be prohibitive. Consumer-grade cameras with wide field of view ("360 cameras") combined with embedded inertial measurement units (IMUs) provide a cost-effective alternative. To support change detection and progress monitoring, highly accurate visual Simultaneous Localization and Mapping (SLAM) and floor plan-referenced localization systems are required. In this paper we present a high-quality dataset collected at an active construction site, which captures realistic challenges such as variable lighting conditions, moving workers, fast motions, and repetitive structures. The dataset offers thirty visual-inertial sequences recorded across seven floors over an eight-month period of the construction project. Ground truth trajectories were collected using a high quality LiDAR-inertial SLAM system rigidly attached to the 360 camera. Additionally, we report the results of an open research challenge evaluating the best visual SLAM and localization systems from around the world. The Challenge attracted substantially higher participation in SLAM, with 62 teams compared to 22 in floor-plan-referenced localization, reflecting the broader maturity of SLAM methods. The higher errors in localization further highlight the difficulty of this task in construction and point to the need for continued research, which this dataset is intended to support. The dataset and the benchmark are publicly available at: https://hilti-trimble-challenge.com/dataset-2026.
Samuele Centanni, Yuhao Zhang, Yifu Tao +8
Hilti AG, Corporate Research & Technology, Schaan, Liechtenstein · University of Oxford, Dynamic Robot Systems Group, Oxford, UK · ETH Zürich, Robotics Systems Lab, Zürich, Switzerland +2
Accurately recovering the camera's calibration and metric poses for any unconstrained video would unlock large-scale training data for navigation and scene understanding. The dominant approaches to this problem are severely limited: Simultaneous Localization and Mapping (SLAM) is sensitive to initialization and transient failures due to its causal, incremental nature; it is often over-optimized for real-time operation and generally requires known camera calibration; while Structure-from-Motion (SfM) typically forgoes any image ordering, enabling optimal initialization and global optimization, but lacks robustness to visual symmetries and extreme motions. To bridge this gap, we introduce a system that combines the strong sequential constraints of SLAM with the flexibility and global optimization of offline SfM, enabling the metric reconstruction of arbitrary, long, uncalibrated videos. This system leverages recent advances in wide-baseline dense image matching, treats temporal ordering as a first-class citizen for reliable loop closure, and augments global optimization with metric monocular depth priors. As a result, thorough evaluations on diverse, challenging datasets that exhibit extreme motion and visual symmetries reveal that our approach is significantly more robust and accurate than both state-of-the-art SLAM and SfM, classical or learned, with given or unknown camera calibration. The code is publicly available at https://github.com/cvg/vidmap.