Built for Denial. Silent in Execution.
Passive, vision-based localization for unmanned ground vehicles. Match vehicle camera imagery against georeferenced Vantor™ global, mission-grade 3D terrain data, then feed position corrections into the navigation stack you already use.
In partnership with
The use case
Maintain absolute positioning in DDIL environments
After GPS loss, Niantic Spatial VPS with Vantor VividTM Terrain uses visual matches to re-anchor your vehicle’s absolute position without replacing your navigation stack. On recorded off-road drives spanning over 31 km, with GPS only at the start, it demonstrated a median absolute-position error of 2.3 m.
INS and odometry errors compound over extended movement, threatening position reporting, mission coordination, and human-machine operations. Each successful match re-anchors the reported position in the global frame.
Niantic Spatial VPS with Vantor VividTM Terrain uses only passive sensors. No active sensors, such as LiDAR, are required.
Pose corrections feed into your existing navigation filter, while your autonomy, INS, and odometry systems continue operating as usual. On-vehicle processing supports localization in network-denied environments. The solution leverages the cameras and vehicle-class edge compute already on the platform, including NVIDIA Jetson chipsets.
The reference layer is built from Vantor 3D terrain data, so no prior ground survey is required.
Horizontal accuracy: 3.5 m at CE90 and 4.3 m at CE95
Your platform is our next proving ground.
Tested on 31 km of off-road driving
Off-road testing (TartanDrive 2.0) using a front camera and lidar-inertial odometry. GPS was used once, at the start, to set the initial search area. The system reported a confident position for 74% of the drive time - a high confidence visual fix, or up to 5 seconds of dead reckoning between fixes, using the vehicle's odometry. Accuracy is measured on all returned positions.
Talk to us
We’re seeking design partners to bring the validated combination of Niantic Spatial VPS and Vantor VividTM Terrain onto UGV platforms. If your vehicles operate over long GPS-denied traverses, talk to us. Together, we’ll define what this capability must prove for your mission.
Frequently Asked Questions
01How does Niantic Spatial VPS and Vantor VividTM Terrain correct a vehicle’s position after GPS loss?
Niantic Spatial VPS with Vantor VividTM Terrain starts from a last good GPS fix or an operator-provided position prior. It matches vehicle camera imagery against georeferenced Vantor satellite terrain data and feeds pose corrections into the existing navigation filter. Each successful match re-anchors the position estimate in the global frame.
02Does it replace our autonomy or navigation stack?
No. Niantic Spatial VPS with Vantor VividTM Terrain adds a source of absolute position. Your existing autonomy, INS, odometry, and navigation filter remain in place. The design partnership defines the interface for consuming pose corrections.
03Does it need dedicated cameras or compute?
The capability is designed to use existing vehicle cameras and vehicle-class edge compute, such as NVIDIA Jetson chipsets, without a dedicated sensor. Camera, compute, and interface fit are assessed against the partner’s platform during the technical working session.
04What happens when visual matching is unavailable?
In network-denied environments, match frequency depends on terrain, lighting, and reference coverage; heavy tree canopy may suppress visual matches. Between matches, the vehicle relies on INS and odometry until sightlines and reference conditions allow the system to re-anchor its position estimate.
05What has been demonstrated, and what is still in validation?
Off-road and urban daylight operation have been demonstrated on recorded drives. Live runs fully on the vehicle over drives of 10 km or more, and night operation in lit urban environments, are in validation. Night operation without ambient light, using IR query imagery, is a future capability.
06What does becoming a design partner involve?
Start with a working session between engineering teams to assess the localization gap and define what an evaluation must prove. A bounded technical evaluation would then use jointly agreed terrain classes, accuracy thresholds, interfaces, test conditions, and exit criteria. Partners provide representative platform access and engineering time to define the integration.