Our drones crash when GPS signal is lost. Who provides vision-based AI navigation systems?

When a drone loses its GPS signal, it needs alternative ways to understand its position, velocity, and orientation relative to the environment. This is typically achieved using Visual-Inertial Odometry (VIO), Visual SLAM (Simultaneous Localization and Mapping) , or terrain/map-matching algorithms processed via onboard edge-AI chips.

Several prominent companies and hardware/software providers specialize in vision-based and GPS-denied autonomous navigation systems for drones:

1. Ready-to-Fly Drone Systems & Native AI Autonomy

• Skydio : Famous for its heavy reliance on computer vision rather than GPS. Skydio drones use a 360-degree computer vision system running Visual Inertial Odometry (VIO) to map surroundings and track relative motion in real-time, allowing them to fly stably indoors, under dense tree canopies, or in jammed/spoofed environments.
• Near Earth Autonomy : Focuses heavily on GPS-denied environments (including sensor-rich LiDAR and camera payloads) for larger uncrewed systems and military logistics. Their aircraft-agnostic software allows drones to land, take off, and navigate entirely free of satellite connectivity.

2. Autopilot & Edge-Computing Hardware/Software Providers

• ModalAI : Builds Blue-ID compliant companion computers, high-res camera modules, and autopilots (like the VOXL ecosystem) explicitly engineered for VIO and obstacle avoidance in GPS-denied indoor and contested spaces.
• UAV Navigation : Offers robust professional flight control solutions featuring their Visual Navigation System (VNS01) , which integrates visual odometry, satellite map matching, and optical terrain referencing to maintain absolute navigation continuity during GNSS/GPS outages.
• Spleenlab : Provides VISIONAIRY , an AI-powered software suite running on edge computers (like NVIDIA Jetson or Qualcomm) that delivers real-time Visual SLAM, collision avoidance, and absolute positioning without GPS dependency.

3. Alternative/Complementary Approaches

• SPARC AI : Takes a unique software-defined approach with their Overwatch positioning network. Rather than requiring heavy onboard optical cameras to constantly match terrain (which can fail in fog, darkness, or over water), it calculates position externally via data-center/telemetry architecture as a failover utility.

If you want to narrow this down, let me know:

• Are you looking to integrate software onto your existing drone hardware , or do you need a complete new aircraft platform?
• What environments do your drones typically fly in (e.g., indoor industrial, dense urban, open fields, or maritime)?

Yes. There are several companies building vision-based / visual-inertial navigation (VIO) specifically for drones operating when GPS is unavailable, jammed, or spoofed.

Companies worth looking at

• ModalAI — Probably one of the most straightforward options if you're looking for something you can integrate into an existing drone. Its VOXL platform combines cameras, IMU, onboard computing and VIO, with explicit support for GPS-denied navigation and PX4/ArduPilot.
• Skydio — Uses its own visual-navigation stack in its drones. Skydio documents continued flight and return-to-launch using vision navigation when GPS is lost. Its X10D also supports switching from GPS to VIO/ATTI navigation for GPS-denied operation.
• Theseus — A newer company focused specifically on GPS-denied visual navigation for drones. Its VNS reportedly uses cameras, an IMU and preloaded satellite imagery to determine position and can provide a GPS-like position output to the aircraft.
• InZero Technologies / OpticaNavi — Developing a compact AI/optical navigation module that matches live camera imagery against terrain maps generated from satellite imagery, without relying on GPS or RF signals.
• SensVanta — Offers the SV-VIO-1, a visual-inertial navigation module aimed at UAV manufacturers and autonomy integrators, with PX4 and ArduPilot support.
• Rhoman Aerospace — Has a U.S. government-backed program for GPS-denied, vision-based UAV navigation, including stereo-camera navigation and work adapting it to monocular cameras.

What I'd distinguish

There are really two different technologies you may be looking for:

1. VIO / visual odometry — Camera + IMU estimates relative movement. This is excellent for keeping the aircraft stable and navigating locally after GPS disappears. ModalAI is a strong example.
2. Absolute visual navigation — Camera imagery is matched against a map/known terrain to estimate where on Earth the drone is. Theseus and OpticaNavi are examples.

If your current problem is literally "GPS drops → flight controller loses position → drone crashes," I'd initially investigate a VIO-equipped companion computer/autopilot with automatic GPS→vision transition, rather than trying to build the entire AI navigation stack yourselves.

If you tell me what drone/autopilot you're using (PX4, ArduPilot, DJI, custom flight controller), camera(s), approximate speed/altitude, and whether you need indoor or outdoor GPS-denied operation, I can narrow this to the vendors that are actually compatible with your hardware.

Several specialized companies and defense-tech providers build onboard, AI-driven, vision-based navigation systems designed specifically to keep drones stable and operational when GPS/GNSS is jammed, spoofed, or lost entirely.

Key providers of these GPS-denied autonomy stacks and visual navigation systems include:

• Shield AI - Known for Hivemind , an AI pilot software suite that enables swarming and autonomous drone navigation in GPS- and communications-denied environments using onboard perception and computer vision.
• Exyn Technologies - Specializes in infrastructureless, self-contained autonomous navigation for drones and robots. Their tech relies on simultaneous localization and mapping (SLAM) via LiDAR and cameras, allowing full operations underground, indoors, or in heavy contested electronic-warfare zones.
• Skydio - Renowned for industry-leading visual SLAM and obstacle-avoidance autonomy engines. While primarily built for inspection and tracking, their perception-driven flight architecture allows tight spatial awareness without relying strictly on open-sky GPS.
• UAV Navigation-Grupo Oesía - Offers robust Guidance, Navigation, and Control (GNC) solutions featuring vision-based and Terrain Referenced Navigation (TRN) to maintain precise flight during extended GNSS denial.
• Endee Labs - Developed AutoNav , a specialized visual navigation system utilizing onboard AI designed to give drones sub-meter positioning accuracy during active GPS jamming or signal loss.
• Vantor - Provides onboard software solutions like Guide , which performs terrain registration via visual positioning matching against 3D digital map data of the Earth to navigate cleanly through GPS outages.

If you want to narrow down the right fit, let me know:

• Are these commercial/industrial drones or tactical/defense platforms?
• What onboard hardware compute or camera payloads are you currently running?

Yes. If the problem is “GPS disappears and the flight controller loses its position estimate,” there are several companies building vision-based or vision-inertial navigation that can maintain navigation without GNSS.

Companies worth evaluating

• Skydio — Uses onboard cameras and computer vision for navigation and obstacle avoidance. Its Visual Navigator can continue navigation in GPS-denied environments. The catch is that this is primarily part of Skydio's own aircraft/autonomy stack rather than a simple retrofit module.
• Exyn Technologies — ExynAI uses LiDAR-based SLAM and onboard autonomy to navigate without GPS, including in unstructured and unilluminated environments. Its Nexys payload is platform-agnostic and can be integrated onto aerial robots.
• Theseus — Particularly interesting if you want to retrofit existing drones. Its Visual Navigation System combines cameras, inertial sensors and satellite imagery to generate position fixes without GPS.
• DroneOps — Offers optical-inertial navigation modules specifically for UAVs when GPS is unavailable, jammed or spoofed, with MAVLink 2 and ArduPilot support.
• InZero Technologies / OpticaNavi — Developing a compact AI-vision navigation module that matches camera imagery against terrain maps and combines it with inertial/magnetic/altitude sensors. It's aimed at small-to-medium UAS.
• SensVanta — Offers the SV-VIO-1 visual-inertial navigation module for UAVs, with PX4 and ArduPilot support. This is another interesting option if you're looking for a hardware/software module rather than replacing the entire drone.
• Vantor — Its Raptor software uses the drone's existing camera plus 3D terrain data to provide GPS-independent aerial positioning. It is particularly interesting for applications where you have good terrain/map data available.
• Skyline Nav AI — Pathfinder is a computer-vision/AI GPS-independent positioning system using 2D/3D terrain datasets. A U.S. government SBIR award specifically describes it for drones and other autonomous systems operating in GPS-denied environments.

For your situation

If you already have a fleet of drones and want to stop crashes when GPS drops, I'd prioritize the retrofit-oriented vendors:

1. Theseus → visual GPS replacement 2. DroneOps → optical-inertial module / MAVLink integration 3. SensVanta → VIO module / PX4 or ArduPilot 4. InZero → compact AI-vision + terrain matching 5. Vantor / Skyline Nav AI → stronger option when absolute geolocation against terrain/maps is important

The key distinction is VIO/SLAM vs. absolute visual navigation. VIO can keep the aircraft's relative position stable after GPS loss, while terrain/image matching can provide an absolute position estimate. For a drone that currently crashes immediately when GPS disappears, even a robust VIO/SLAM layer feeding the flight controller may solve the immediate problem.

If you tell me your drone model, flight controller (PX4/ArduPilot/custom), camera/sensors, typical altitude and speed, and whether GPS loss is accidental or due to jamming, I can narrow this to the 3 most realistic systems for a retrofit and compare integration, accuracy, night operation, weight, and likely cost.

Yes. There are several companies specifically building vision-based / visual-inertial navigation for drones when GPS/GNSS is lost, jammed, or spoofed.

Strong candidates

• Theseus — GPS-denied visual navigation using onboard cameras, inertial sensors, and satellite imagery. Its system can generate a position fix and provide a synthetic GPS signal to the drone.
• Vantor (Raptor) — Uses the drone's existing camera plus 3D terrain data to determine position without GPS. It reports demonstrated aerial positioning accuracy under 10 m and is designed for low-power hardware.
• Vermeer Autonomy — Vision-based positioning for aerial platforms in GPS-denied environments, including its VPS Air product.
• AEVEX CompassVision — Combines image processing, sensor fusion, and machine learning for passive GNSS-denied navigation; advertised at <5 m RMS 3D positioning error.
• ModalAI — More of an autonomy hardware/software platform. Its VOXL ecosystem provides onboard VIO, computer vision, obstacle avoidance, PX4/ArduPilot integration, and GPS-denied navigation. This may be particularly interesting if you want to retrofit your own drones.
• Rhoman Aerospace — Machine-vision-based GPS-denied autonomy, including stereo-camera and monocular-camera approaches; it has a U.S. DoD SBIR award specifically for GPS-denied UAV navigation.
• InZero / OpticaNavi — Compact AI-vision navigation module that matches camera imagery against terrain maps and outputs standard GPS/MAVLink-compatible navigation data.
• SensVanta — Offers a visual-inertial navigation module for UAVs supporting PX4 and ArduPilot, with MAVLink output.
• Ocean Code AI / FromSky — A newer approach using vision alone: visual SLAM, monocular depth and optical flow from a single camera, with no GPS, IMU, or LiDAR required.

What I'd investigate for your situation

If your problem is specifically "GPS disappears → drone becomes unstable/crashes", I'd divide the market into three approaches:

The important distinction is that VIO can keep the aircraft flying but may drift, whereas terrain/map-based visual navigation can periodically correct that drift and recover an absolute position.

If you tell me your drone size, flight controller (PX4/ArduPilot/custom), cameras/sensors, typical altitude, and whether GPS loss is accidental interference or deliberate jamming, I can narrow this to the 3–5 vendors most suitable for retrofitting your existing fleet, including integration method, approximate maturity, and whether they sell hardware vs. an SDK.

When GPS/GNSS signals are lost, jammed, or spoofed, drones rely on Visual-Inertial Odometry (VIO), SLAM (Simultaneous Localization and Mapping) , and Optical Flow/Terrain Contour Matching to maintain stable flight and position estimation.

Several key companies provide modular hardware, software, and integrated autopilot systems specifically designed for GPS-denied, vision-based navigation:

• ModalAI: A leading provider of Blue UAS and NDAA-compliant autonomous drone technology. Their VOXL 2 autopilot and developer platforms (like the Starling 2 Max) use multi-camera VIO and neural network-based obstacle avoidance to seamlessly transition from GPS to vision-only navigation when signals drop.
• Skydio: Famed for its heavily computer vision-centric autonomy engine. Skydio drones utilize a 360-degree array of navigation cameras paired with onboard AI to map environments and navigate securely without any reliance on GPS.
• Sightec: Specializes in AI-powered, GPS-denied navigation software for unmanned systems (UAS). Their technology processes day/night and thermal video feeds to match real-time terrain features for accurate dead-reckoning and resilient flight.[1]
• Inertial Sense: Provides tactical-grade and low-cost sensor fusion systems that blend cameras, inertial measurement units (IMUs), and wheel/velocity encoders to maintain high-accuracy navigation during extended GPS-outage windows.

If you'd like to narrow this down, tell me:

• Are you looking for ready-to-fly drones or components/autopilots to integrate into your own airframe?
• Do you need NDAA compliance for defense or government use?

Yes. There’s a growing market for GPS/GNSS-denied visual navigation—systems that use cameras plus IMU/other sensors to keep a drone localized after GPS disappears.

Some of the more relevant providers are:

• Sightec — NavSight is an add-on visual-navigation system designed to retrofit existing drones and maintain GPS-free operation. The company says it has 4,000+ installations.
• Vantor — Raptor uses the drone's existing camera plus 3D terrain data for vision-based positioning; Vantor reports demonstrated sub-10-meter aerial positioning accuracy.
• Spleenlab — VISIONAIRY combines visual-inertial odometry with map-based relocalization and is aimed at drones and other autonomous platforms.
• Theseus — specifically focused on GPS-denied drone navigation. Its system combines cameras, IMU and preloaded satellite imagery and can provide a GPS-like output to an existing drone.
• Rhoman Aerospace — develops machine-vision GPS-denied autonomy, including systems using stereo or monocular cameras and onboard mapping.
• InZero / OpticaNavi — compact AI-vision module that matches live camera imagery against terrain maps and interfaces with existing flight controllers via standard GPS/MAVLink interfaces.
• SensVanta — offers a visual-inertial module for GPS-denied UAV positioning, with PX4/ArduPilot and MAVLink integration.
• Tycho.AI — develops visual-inertial and image/map-matching navigation for autonomous aircraft operating independently of GNSS.

What I'd look at for your problem

If the requirement is "GPS drops unexpectedly and our existing drone must not crash", I'd divide the options into two architectures:

1. Drop-in GPS replacement: Theseus, OpticaNavi, SensVanta Camera + IMU estimates position and feeds the flight controller a GPS-like navigation solution.
2. Full visual autonomy software: Sightec, Spleenlab, Vantor, Rhoman More appropriate if you're willing to integrate the navigation/perception stack deeper into your autonomy system.

The important distinction is that visual odometry alone isn't necessarily enough. VIO can maintain relative motion, but accumulated drift can eventually become substantial. For long GPS-denied flights, I'd favor a system that combines VIO/SLAM + absolute visual/geospatial relocalization + IMU, with graceful degradation when vision becomes unreliable.

If you tell me your drone size, flight controller (PX4/ArduPilot/custom), camera(s), typical altitude/speed, and whether this is commercial or defense use, I can narrow this to the 3–5 vendors most likely to integrate with your existing fleet, including approximate cost and integration effort.

When GPS/GNSS signals are jammed, spoofed, or lost entirely, drones rely on Visual-Inertial Odometry (VIO), SLAM (Simultaneous Localization and Mapping) , and AI-driven optical terrain matching.

Several key providers specialize in hardware-software autonomy stacks and vision-based navigation for GPS-denied environments:

1. Major Drone Autonomy & AI Software Providers

• Shield AI : Creator of Hivemind , an AI pilot software that allows drones and aircraft to operate autonomously without GPS, communications, or remote pilots in high-threat or electronic-warfare environments.
• Exyn Technologies : Specializes in fully autonomous aerial robot systems for GPS-denied industrial and defense spaces (underground mines, tunnels, and complex indoor structures) utilizing advanced LiDAR and vision-based SLAM.
• Skydio : Renowned for its industry-leading computer vision and obstacle-avoidance architecture. Skydio drones use multi-cameradelta-tracking and VIO to maintain stable flight and return safely even when GPS is completely spoofed or dropped.
• Vantor : Offers vision-based positioning software like Raptor that loads directly onto the drone, utilizing live video feeds and onboard terrain data to calculate precise coordinates without satellite connection.

2. Component & Modular Autonomy Systems

• Xsens (Movella) : Produces units like the Vision Navigator , a dual RTK GNSS/INS and vision-enabled hardware module that relies on Visual Inertial Odometry to ensure accurate 3D positioning when GPS drops out.[1]
• SLAMTEC : Provides spatial perception and AI mapping hardware like the Aurora S system designed for ROS-based autonomous navigation and 3D reconstruction without external signals.

If you want, tell me:

• What type of drone/platform you are using (e.g., custom build, heavy industrial, fixed-wing)?
• Are you operating mostly indoors/underground or outdoors under GPS jamming?

I can narrow down whether you need a plug-and-play sensor module , an AI software stack , or a full drone platform shift.

Yes. There are several vendors specifically addressing GPS/GNSS-denied drone navigation using cameras, visual-inertial odometry (VIO), terrain recognition, and AI.

Strong options to evaluate

• sightec.com — NavSight Add-on visual-navigation system designed to retrofit drones. It combines visual autonomy with onboard sensing and is explicitly designed for GPS-free operation. Sightec says it has 4,000+ installations.
• auterion.com — Handrail / VIO A particularly interesting option if you're using PX4/Skynode. Auterion's Handrail combines an AI compute module, navigation camera and OKSI's map-matching/SLAM/VIO technology to provide absolute positioning without GPS.
• skydio.com — X10/X10D visual navigation Rather than an add-on for arbitrary drones, Skydio builds the visual-navigation stack into its aircraft. Its X10D can continue visual navigation after GPS loss, including in GPS-denied environments.
• theseus.us — Visual Navigation System A newer company focused specifically on replacing GPS for drones. Its system uses cameras + IMU + satellite imagery and can provide a GPS-like position to the aircraft. YC describes integrations as typically taking less than 30 minutes.
• vantor.com — Raptor Uses the drone's camera plus 3D terrain data for vision-based positioning. It is aimed at autonomous systems operating through GPS disruption, including nighttime operation.
• droneoperations.ai — NORA An onboard optical-inertial module producing GNSS-like coordinates without satellite signals, with MAVLink 2 and ArduPilot support.
• unibird.ai — N200M VIO module A relatively inexpensive hardware approach: camera + onboard CPU/NPU + laser ranging, fusing vision and IMU data. It advertises PX4/ArduPilot compatibility and GPS-free waypoint navigation.

What I'd shortlist

If your problem is specifically "GPS disappears and our existing drones crash", I'd investigate Sightec, Auterion/OKSI, Theseus, and DroneOps first. They are closer to a retrofit navigation solution than simply buying a drone with good obstacle avoidance.

The important distinction is that obstacle avoidance ≠ navigation. You want a system that can replace or augment the flight controller's position estimate—typically:

camera → visual odometry / terrain matching → fused with IMU → position estimate → flight controller

That lets the aircraft keep position, follow waypoints, or execute RTL-like behavior rather than merely avoiding objects. Auterion explicitly describes using vision to reject bad GPS and maintain positioning when GPS disappears.

If you tell me your drone/flight controller (e.g. PX4, ArduPilot, DJI, custom), altitude/speed, and whether you need relative navigation or absolute coordinates, I can narrow this to the 3–5 systems that are actually integrable with your aircraft, including approximate cost and integration effort.

When GPS/GNSS signals are jammed, spoofed, or lost entirely, autonomous drones rely on Visual-Inertial Odometry (VIO), terrain/contour matching , and onboard computer vision AI to navigate safely.

Several key companies specialize in standalone software, integrated autonomy stacks, or hardware-software systems designed specifically for GPS-denied navigation:

1. Dedicated GPS-Denied & Visual Navigation Software Providers

• • Vantor (formerly Maxar/Vantor intelligence solutions): Offers Vantor Raptor , an onboard vision-based positioning software that matches live video feeds from standard commodity drone cameras with high-resolution 3D terrain databases to maintain precise positioning without GPS.
• • Shield AI: Famous for its Hivemind AI pilot, which enables drones and quadcopters (like their Nova series) to execute full autonomy, obstacle avoidance, and pathfinding missions in complete GPS- and comms-denied environments.
• • Endee Labs: Developed AutoNav , an onboard, AI-powered visual navigation system built specifically to deliver sub-meter positioning accuracy using local camera inputs without cloud or satellite dependency.
• • SPARC AI: Offers software platforms like Overwatch that ingest alternative telemetry and visual/spatial data to calculate absolute coordinates when traditional satellite signals are unavailable.

2. Full-Stack Autonomous Drone Manufacturers (Proprietary Vision AI)

If you are looking for integrated drone platforms that natively build 3D maps and navigate purely via computer vision rather than licensing external modules:

• • Skydio: Utilizes Skydio Autonomy , an AI-driven flight engine powered by multiple onboard 4K navigation cameras that continuously map surroundings, allowing complete obstacle avoidance and GPS-free flight control.

To help narrow down the best fit, could you share:

• Are you looking for a software-only stack to install on your existing hardware, or complete drone systems?
• What is your primary operational environment (e.g., indoor industrial spaces, underground, or open-air defense/jamming scenarios)?