For fixed-wing aircraft, reliable localization during approach and landing is critical to flight safety. Visual--inertial navigation is promising, but aircraft navigation also requires integrity: trustworthy outputs and timely warnings when they should not be used. Real approach data are costly to collect, while many visual errors and failures required for integrity algorithm development are difficult to reproduce safely. We present PASS, a trustworthy Precision Approach Simulation System for CAT III integrity algorithm development. In a unified temporal and geometric framework, PASS integrates configurable airport environments, prescribed trajectories, inertial simulation, calibrated stereo images, and continuous runway-relative ground truth. It supports navigation and integrity research through controllable sensor errors and environmental variations, computable visual projection errors, deterministic runway geometry priors, and condition-labeled data for statistical error modeling. We further propose Sim-to-Real Scene Trustworthiness (SRST), a measure based on DINOv2 feature distributions and UMAP, to quantify sim-to-real closeness. Experiments with representative stereo--inertial estimators demonstrate the usability of the generated sequences and quantify localization errors across algorithms and environments. PASS provides a controllable foundation for developing and evaluating visual--inertial navigation and integrity algorithms for fixed-wing precision approaches up to CAT III requirements.
Figures & tables
Resource
Type
Fixed-wing runway approach
Continuous stereo–IMU
Runway-relative ground truth
Configurable errors or conditions
Integrity support information
Sim-to-real credibility assessment
LARD v1 Ducoffe et al. (2023)
Dataset
Yes
No
No
No
No
Partial
LARD v2 Bougacha et al. (2026)
Dataset
Yes
No
No
No
No
Partial
BARS Chen et al. (2023)
Dataset
Yes
No
No
No
No
No
RLD/VALNet Wang et al. (2024)
Dataset
Yes
No
No
No
No
No
UnrealCV Qiu and Yuille (2016)
Simulation framework
No
No
No
Partial
No
No
AirSim Shah et al. (2017)
Simulation framework
No
Supported
No
Supported
No
No
Table 1: Comparison of representative runway-oriented datasets and synthetic simulation resources for fixed-wing precision approach navigation and integrity research.
Figure 1: Overall architecture and information flow of PASS . The Approach Scenario Module defines the runway geometry, operating conditions, and controllable disturbances required for downstream integrity analysis. The Trajectory and IMU Module generates the prescribed aircraft states, runway-relative ground truth, and simulated inertial measurements. The Stereo Imaging Module replays the prescribed motion and records synchronized stereo images. A shared time base and coordinate transformation chain maintain temporal and geometric consistency across the three modules.
Table 2: Inputs and outputs of the three modules in PASS shown in Fig. 1 .
Figure 2: Coordinate framework of PASS . The transformation chain connects the geodetic and local navigation frames to the runway, aircraft body/IMU, stereo camera, and image frames, providing a common geometric basis for multimodal data generation.
Error component
Parameter(s)
Modeled effect
Sampling configuration
Δtimu,fimu
Integration interval and measurement rate
Constant bias
eb,db
Constant gyroscope and accelerometer offsets
Random walk
web,wdb
Uncorrelated stochastic error accumulation
Correlated bias
R0G,R0A,τG,τA
First-order Gauss–Markov bias process
Scale-factor error
δKG,ii,δKA,ii
Axis-dependent multiplicative error
Axis misalignment
δKG,ij,δKA,ij
Off-diagonal cross-axis coupling
Table 3: Parameterization of the PSINS-based IMU simulation model.
Figure 3: Virtual stereo camera calibration: (a) relative configurations of the checkerboard and camera used for multi-view calibration and (b) distribution of the resulting reprojection errors.
Parameter
Left camera
Right camera
Image resolution (pixels)
1920×1080
fx (pixels)
960.6297
959.6249
fy (pixels)
960.4265
959.4180
cx (pixels)
962.9833
961.6595
cy (pixels)
540.6769
540.3199
k1
1.796×10−4
1.462×10−3
Table 4: Calibrated parameters of the stereo camera system used in PASS . The distortion coefficients follow the order [k1,k2,p1,p2] .
Figure 4: Controlled environmental variations represented in PASS : (a) illumination conditions at different times of day and (b) representative weather conditions. The prescribed aircraft trajectory and runway reference geometry remain unchanged across these variations.
Figure 5: Dynamic-object-induced visual perturbation in the approach view. The birds introduce localized motion and partial occlusion of runway features without modifying the prescribed aircraft trajectory or reference geometry.
Figure 6: Runway geometry and runway-centered reference frame used by PASS . The runway origin, dimensions, corner coordinates, principal line features, and runway heading ψrwy define deterministic geometric references for runway feature projection and error evaluation.
Category
ICAO
Runway
Length (m)
Width (m)
Urban Airport
KLAX
L6–R24
3381.7560
60.9600
Urban Airport
ZSPD
R34–L16
3799.9416
60.0456
Desert Airport
OMAA
R31–L13
4099.8600
60.0456
Mountainous Airport
LOWI
26–08
2000.0976
45.1104
Coastal Airport
LFMN
R22–L40
2627.9856
45.1104
Coastal Airport
LGMK
34–16
1901.9520
29.8704
Table 5: Airport categories included in PASS dataset.
Figure 7: Representative real and simulated approach images from coastal, desert, mountain, and urban airport environments.
Figure 8: Human-inspired decision process of SRST for unpaired real and simulated approach images. The 4c formulation requires the assigned category to prevail over at least two of three alternatives, whereas the 3c formulation merges coastal and urban scenes and requires the assigned category to prevail over both remaining alternatives.
Airport
Estimator
15 s
20 s
25 s
30 s
ZSPD
VINS-Fusion
3.66
7.48
12.65
19.95
ORB-SLAM3
0.78
1.32
2.02
3.02
KSMO
VINS-Fusion
12.45
20.18
30.58
40.51
ORB-SLAM3
7.50
10.66
14.15
18.51 †
LGMK
VINS-Fusion
9.64
13.87
18.40
23.74
ORB-SLAM3
1.40
2.41
3.81
5.48
Table 6: Mean terminal position error [m] for the stereo–inertial terminal approach experiments. Values are in metres and are averaged over the repetitions that reached the planned final frame. Each condition contains three runs unless marked otherwise.
Figure 9: Evolution of the runway-relative position error over the 30 s terminal approach windows. Each panel shows three independent executions under an identical input and configuration. The cross denotes the last valid output of the interrupted KSMO ORB-SLAM3 run and does not represent the planned terminal error. The vertical axes use a logarithmic scale.
Middle approach
Low approach
Category
Real
Simulated
Real
Simulated
Coastal
89
41
62
60
Desert
27
61
34
40
Mountainous
56
59
52
42
Urban
104
79
45
22
Table 7: Real and simulated image counts by approach stage and airport scene category.
Figure 10: Joint two-dimensional UMAP embeddings of 469 real and 404 simulated approach images, fitted separately for the middle and low stages of approach. Colors indicate airport scene categories; circles and triangles denote real and simulated images, respectively. Coordinates are comparable within, but not between, the two panels.
UMAP Test
Overall, n/404
Middle coastal, n/41
Middle urban, n/79
Test 1
364 (90.1%)
20 (48.8%)
60 (75.9%)
Test 2
383 (94.8%)
20 (48.8%)
79 (100.0%)
Test 3
365 (90.3%)
38 (92.7%)
43 (54.4%)
Test 4
368 (91.1%)
20 (48.8%)
64 (81.0%)
Test 5
366 (90.6%)
20 (48.8%)
62 (78.5%)
Table 8: SRST results for five selected UMAP realizations. The six unlisted stage-by-category groups each had a 100% image-level concordance rate in all five realizations.
UMAP test
Overall, n/307
Middle urban/coastal, n/63
Low urban/coastal, n/42
Test 1
307 (100.0%)
63 (100.0%)
42 (100.0%)
Test 2
306 (99.7%)
63 (100.0%)
41 (97.6%)
Test 3
305 (99.3%)
63 (100.0%)
40 (95.2%)
Test 4
304 (99.0%)
63 (100.0%)
39 (92.9%)
Test 5
303 (98.7%)
63 (100.0%)
38 (90.5%)
Mean of 50 tests
98.1%
Table 9: Three-category SRST results. The five illustrated realizations were selected for their high overall scores and do not estimate performance across random initializations. The final row reports the mean across all 50 realizations. The four unlisted stage-by-category groups passed completely in the five illustrated realizations.
Figure 11: Joint two-dimensional UMAP embeddings for one illustrative three-category realization with an overall SRST score of 98.7%. Middle and low approach are shown separately. Colors indicate the merged urban/coastal, desert, and mountainous scene categories; circles and triangles denote real and simulated images, respectively.
Vision-based navigation complements Global Navigation Satellite Systems, but certification demands integrity guarantees that account for faulty measurements. Previous work presented a probabilistic computer vision pipeline for runway-based pose estimation with fault detection inspired by Receiver Autonomous Integrity Monitoring. This work extends that framework by deriving protection levels, which provide probabilistic bounds on pose error that remain valid under undetected faults. We present an algorithm for computing protection levels for the nonlinear Perspective-n-Point problem applied to an aviation setting. The algorithm covers all six degrees of freedom of the aircraft pose (position and orientation) directly. We analyze the effect of measurement redundancy, pixel-level prediction uncertainty, and runway distance on the resulting protection levels. To make the results tangible, we demonstrate tradeoffs in the protection levels on an illustrative runway example.
Olivia Beyer Bruvik, Romeo Valentin, Marc R. Schlichting +2
Department of Aeronautics and Astronautics, Stanford University, Stanford, CA 94305 USA · A3 by Airbus LLC, Sunnyvale, CA 94086 USA
We introduce SCAR, a method for long-term auto-calibration refinement of aerial visual-inertial systems that exploits georeferenced satellite imagery as a persistent global reference. SCAR estimates both intrinsic and extrinsic parameters by aligning aerial images with 2D--3D correspondences derived from publicly available orthophotos and elevation models. In contrast to existing approaches that rely on dedicated calibration maneuvers or manually surveyed ground control points, our method leverages external geospatial data to detect and correct calibration degradation under field deployment conditions. We evaluate our approach on six large-scale aerial campaigns conducted over two years under diverse seasonal and environmental conditions. Across all sequences, SCAR consistently outperforms established baselines (Kalibr, COLMAP, VINS-Mono), reducing median reprojection error by a large margin, and translating these calibration gains into substantially lower visual localization rotation errors and higher pose accuracy. These results demonstrate that SCAR provides accurate, robust, and reproducible calibration over long-term aerial operations without the need for manual intervention.
Henry Hölzemann, Michael Schleiss
Department of Sensor Data and Information Fusion, Fraunhofer FKIE, 53343 Wachtberg, Germany · Chair of Astronautics, University of the Bundeswehr Munich, 85579 Neubiberg, Germany
This paper presents an analysis of the observability and consistency properties of filtering-based visual-inertial navigation systems (VINS) that utilize anchored feature representations. The unobservable subspace of VINS with anchored landmark parameterizations is shown to be independent of the estimated landmark state, which leads to improved estimator consistency properties without any additional modifications. However, the unobservable subspace is still found to depend on the estimated navigation state, necessitating additional consistency-enforcing techniques. Two methods to improve the consistency of VINS with anchored feature representations are presented. Simulation results showcase that all estimators employing anchored feature paramterizations exhibit improved consistency properties compared to algorithms that estimate features resolved in a global reference frame, especially in scenarios where feature initialization may be poor. Real-world experiments on the TUM-VI dataset showcase that the use of anchored feature representations alone can yield comparable performance to consistency-improved estimators employing a global feature representation, demonstrating the benefit of using anchored feature parameterizations for VINS.
Mitchell Cohen, Vassili Korotkine, James Richard Forbes
Department of Mechanical Engineering, McGill University, Montreal, QC H3A 0C3, Canada