Support
Contact Us
Contact Us
White Papers
All White Papers

VINS GNSS-Denied Navigation Performance Across Multiple Flight Tests

Eighteen real flights, zero GPS. See how VINS maintained sub-20 m accuracy over 150 km using visual-inertial sensor fusion
This case study summarizes a comprehensive flight test campaign conducted to evaluate the performance of the Visual Inertial Navigation System (VINS) using Visual Positioning System (VPS) and Visual Inertial Odometry (VIO) aiding during GNSS-denied operations. The tests were executed on a Cessna 172 aircraft using both daylight and infrared cameras to assess VINS GNSS-denied navigation performance under diverse visual conditions and terrains, including mountainous, riverine, and agricultural areas.

The Challenge

In GNSS-denied environments, conventional navigation systems relying solely on inertial sensors experience rapid error accumulation due to uncorrected drift. This results in unreliable position estimates and limits mission continuity. The challenge addressed in this campaign was to demonstrate that VPS and VIO could maintain stable, accurate GPS-denied navigation across extended flight segments over distances up to 150 km — without a single satellite fix.

The Solution

The VINS integrates visual map matching (VPS) with VIO in a tightly coupled sensor fusion architecture. Real-time imagery captured by the day and infrared cameras is compared against onboard 3D satellite maps to generate absolute position updates. These updates are fused in an Extended Kalman Filter to correct the INS (Inertial Navigation System), significantly reducing drift. When VPS confidence drops, VIO maintains continuity — making this a robust solution for autonomous navigation without GPS.

The Method

A total of 18 flight datasets were collected, with GNSS intentionally disabled during segments of up to 46 minutes and distances up to 150 km. Test altitudes ranged from 200 to 2,000 meters AGL. The flights included diverse maneuvers and environmental conditions to challenge the system, ensuring realistic evaluation across customer use cases in defense and commercial aviation.

Figure 1. Flight Path Trajectories
Figure 1. Flight Path Trajectories

The Results

Table 1.
Flight ID Position Error (m) Max AGL (m) Avg AGL (m) Flight Duration (min) GNSS-Denied Time (min) GNSS-Denid Distance (km)
2025-04-28-F1
14.32
1070.37
602.59
30.88
23.55
69.73
2025-04-28-F2
8.70
856.33
586.25
28.59
20.22
48.50
2025-04-28-F3
12.17
981.43
558.54
23.71
19.06
56.77
2025-06-05-F1
18.31
2078.25
1024.42
49.81
44.13
125.56
2025-06-05-F2
7.15
346.17
256.36
22.07
15.36
40.26
2025-06-21-F1
17.60
362.28
227.88
40.23
22.09
69.77
2025-06-21-F2
8.08
363.63
229.88
40.03
21.88
69.53
2025-06-21-F3
23.42
350.45
188.95
22.07
15.49
45.56
2025-06-21-F4
9.07
350.26
190.10
21.99
15.10
44.70
2025-06-25-F1
9.38
339.39
251.18
34.33
31.19
93.68
2025-07-10-F1
7.81
667.33
316.07
32.09
24.84
77.77
2025-07-10-F2
9.92
570.50
329.27
22.74
18.90
52.84
2025-07-10-F3
14.78
999.12
555.89
22.79
21.84
62.15
2025-07-10-F4
35.09
999.79
2558.91
22.68
21.79
62.36
2025-07-22-F1
14.72
1285.08
922.00
36.81
28.32
101.15
2025-07-22-F2
11.38
1286.80
924.53
36.81
28.35
101.31
2025-07-22-F3
25.28
789.48
395.72
47.41
46.28
152.61
2025-07-24-F1
29.60
712.23
377.74
38.61
28.88
102.23

The Discussion

Across all flights, the VINS system achieved a mean position error of 16.55 meters, demonstrating highly stable GNSS-denied navigation over extended operations. The infrared camera (mean error 15.88 m) consistently outperformed the daylight camera (mean error 18.08 m), providing improved tracking under low-light and complex environmental conditions.
Figure 2. Probability Distribution of Error
Figure 2. Probability Distribution of Error
The system also achieved a CEP50 of 12.67 meters over a 50 km trajectory, meaning that 50% of all position estimates remained within a 12.67-meter radius. This highlights the solution’s ability to maintain a tightly bounded error region even during long-duration, GPS-denied flights.
Figure 3. CEP50 at 50 km
Figure 3. CEP50 at 50 km
Performance analysis showed that position error per distance traveled remained below 0.2% for most flights, and total accumulated error never exceeded 0.5% over 100 km. These results confirm the robustness of the multi-aiding architecture, ensuring mission continuity even when visual conditions degrade.
Figure 4. Error vs Distance traveled
Figure 4. Error vs Distance traveled

The Benefits

Business Impact:
  • Reliable visual-inertial navigation over 150 km without GNSS
  • Sub-20 meter average error across diverse terrain and lighting conditions
  • CEP50 of 12.67 m ensures tightly clustered position estimates
  • Scalable for defense and commercial aviation applications

Explore Inertial Labs’ related solutions:

Ready to navigate without GPS?

Contact the Inertial Labs team to discuss VINS GNSS-denied navigation for your platform and mission requirements.

Tel.: +1 (703) 880-4222
E-mail: il.sales@viavisolutions.com
Adresse: 39959 Catoctin Ridge St, Paeonian Springs, VA 20129, USA
inertiallabs.com

Latest White Papers

All White Papers
Scroll to Top

Website maintenance has been scheduled for Sunday, April 2 from 7 am to 9 pm EDT.
The resource may be unavailable at this time. Please accept our apologies for any inconvenience.