GNSS-Denied Drone Navigation

By Association for Drones

Published 23 September 2026

Global Navigation Satellite Systems such as GPS, Galileo, GLONASS and BeiDou have become fundamental to modern drone operations. They provide aircraft with position, velocity and timing information that supports navigation, automated flight, mapping, georeferencing and return-to-home functions. However, there are many environments where satellite navigation is unavailable, unreliable or insufficiently accurate.

GNSS-denied drone navigation refers to the technologies and techniques that allow an unmanned aircraft to determine its movement and position without depending continuously on satellite navigation. Instead, the drone can combine information from cameras, inertial sensors, LiDAR, radar, optical-flow sensors, terrain data and other onboard systems to estimate where it is and how it is moving.

This capability is increasingly important for indoor inspection, underground mining, tunnels, warehouses, industrial facilities, infrastructure inspection, urban environments, forestry, emergency response and other locations where satellite visibility may be poor.

GNSS-denied navigation is also becoming an important foundation for more autonomous drones. Rather than depending on one positioning technology, future aircraft are increasingly likely to combine several navigation sources and automatically determine which information can currently be trusted.

However, GNSS-denied navigation should not be interpreted as perfect positioning without GPS. Every alternative navigation technology has limitations. Visual systems require useful visual features, LiDAR requires measurable geometry, inertial systems accumulate drift and map-based navigation depends on the quality of the reference data.

The strongest systems therefore use multi-sensor navigation, combining complementary technologies to maintain a reliable estimate of the aircraft’s position.

Why Drones Normally Depend on GNSS

Most commercial drones use GNSS as one of their primary navigation sources. When sufficient satellite signals are available, the aircraft can estimate its geographic position and use that information to hold position, follow waypoints and return to a recorded location.

Professional mapping drones may additionally use Real-Time Kinematic or Post-Processed Kinematic GNSS to improve positioning accuracy.

GNSS is extremely effective because the positioning reference exists independently of the drone. The aircraft does not need to understand its surroundings to determine its approximate global position.

However, satellite signals arriving at Earth are relatively weak. Buildings, terrain, vegetation and other structures can block or degrade reception. Reflections can also introduce errors.

This means a drone designed to operate beyond open outdoor environments needs additional navigation technologies.

What Does GNSS-Denied Mean?

A GNSS-denied environment is one in which reliable satellite positioning cannot be continuously obtained.

The most obvious examples are indoor and underground locations. Satellite signals generally cannot penetrate sufficiently through buildings, tunnels, mines or substantial underground structures.

Other environments may be GNSS-degraded rather than completely denied. A drone operating between tall buildings, beneath a bridge, under dense vegetation or close to large structures may still receive satellite signals, but their quality may fluctuate.

The distinction is important because some navigation systems can use GNSS whenever it is available and transition toward other sensors when its reliability decreases.

This hybrid approach can be more robust than treating navigation as either GNSS or non-GNSS.

The Importance of Sensor Fusion

GNSS-denied navigation rarely depends on one sensor.

Instead, modern systems combine several measurements through sensor fusion.

An IMU can measure rapid movement but gradually drifts. Cameras can observe environmental features but may struggle in darkness. LiDAR works without ambient light but may struggle in certain geometrically repetitive environments. Radar can operate in difficult visibility but may provide different levels of spatial detail.

Combining these sources allows one sensor to compensate for limitations in another.

The navigation computer continuously estimates the aircraft’s state from these observations.

This can include position, velocity, orientation and sometimes the uncertainty associated with those estimates.

A sophisticated navigation system should not only calculate where it believes the aircraft is but also estimate how confident it is in that calculation.

Inertial navigation is one of the foundations of GNSS-independent flight.

An Inertial Measurement Unit typically contains accelerometers and gyroscopes that measure acceleration and rotational movement.

These measurements allow the navigation system to estimate how the aircraft has moved from a previous position.

The major advantage is independence from external signals. An IMU continues operating indoors, underground and in darkness.

However, small measurement errors accumulate.

This is known as inertial drift.

Over short periods, inertial navigation can provide excellent motion information. Over longer periods, the estimated position can increasingly differ from the aircraft’s actual location.

For this reason, inertial navigation is normally combined with other sensors capable of correcting accumulated drift.

IMU Quality

Not all IMUs provide the same performance.

Consumer drones may use relatively low-cost MEMS inertial sensors, while professional navigation systems can use higher-performance inertial technology.

Better sensors generally accumulate error more slowly, but size, weight, power consumption and cost can increase.

For drone applications, the objective is usually not to eliminate inertial drift entirely. Instead, the system combines a suitable IMU with external observations that regularly constrain the accumulated error.

The IMU remains extremely important because it provides high-rate motion information between those corrections.

Visual Odometry

Visual odometry estimates drone movement by analysing consecutive camera images.

Software identifies features within an image and tracks how those features move between frames.

From this apparent motion, the system estimates how the camera has moved.

Buildings, corners, rocks, machinery and textured surfaces can provide useful visual features.

Visual odometry can work extremely well in environments containing adequate lighting and distinctive objects.

However, it may struggle with darkness, fog, smoke, repetitive patterns, featureless walls or rapidly changing scenes.

It therefore works best as part of a broader navigation architecture.

Visual-Inertial Odometry

Visual-Inertial Odometry, commonly abbreviated VIO, combines camera information with IMU measurements.

The IMU provides rapid motion estimates while cameras provide environmental references that help constrain drift.

This combination has become increasingly important for autonomous drones and robotics.

VIO can support flight inside buildings, warehouses and other environments where GNSS is unavailable.

However, visual conditions remain important.

Poor lighting, motion blur, transparent surfaces or areas with very little visual texture can reduce performance.

The navigation system should therefore understand when visual confidence is deteriorating rather than blindly trusting every camera observation.

Optical Flow

Optical-flow sensors measure apparent movement across an image.

Many consumer and professional drones use downward-facing cameras for this purpose.

If the ground texture moves across the image, the system can estimate relative movement.

Optical flow is particularly useful for maintaining stable position when GNSS is unavailable.

Indoor drones frequently use it together with altitude sensors.

However, optical flow is usually a relative positioning method.

It does not automatically tell the aircraft its global geographic position.

Performance can also degrade over uniform surfaces such as featureless floors, reflective water or surfaces with repeating patterns.

LiDAR provides another powerful approach to GNSS-denied navigation.

The sensor measures distances to surrounding surfaces and creates a three-dimensional representation of the environment.

As the drone moves, software compares new LiDAR measurements with previously observed geometry.

This allows the system to estimate movement.

LiDAR is particularly valuable in darkness because it is an active sensor and does not depend on ordinary visible illumination.

Applications include mines, tunnels, industrial facilities and indoor infrastructure.

However, LiDAR navigation can still struggle in environments containing limited or repetitive geometry.

LiDAR-Inertial Odometry

LiDAR-Inertial Odometry combines LiDAR measurements with IMU information.

The IMU predicts the drone’s movement while LiDAR scans provide geometric constraints.

This can provide robust localisation in environments where cameras struggle.

For example, an underground tunnel may be completely dark but still contain enough structural geometry for LiDAR localisation.

However, long uniform tunnels can create challenges because the geometry changes very little along the direction of travel.

The strongest systems therefore combine LiDAR, inertial measurements and additional navigation references where practical.

SLAM

Simultaneous Localization and Mapping, or SLAM, allows a drone to build a map while estimating its own position within that map.

This is particularly useful when the aircraft enters an environment that has not previously been mapped.

The drone observes its surroundings using LiDAR, cameras or both.

The navigation system identifies relationships between successive observations and estimates the aircraft’s movement.

At the same time, those observations are added to an expanding environmental map.

SLAM therefore provides both navigation information and a useful mapping product.

Loop Closure

SLAM systems can accumulate positional drift as the drone travels.

Loop closure helps reduce this.

When the aircraft returns to a previously mapped area, the system recognises familiar geometry or visual features.

It then knows that the two estimated locations should correspond.

This allows accumulated error to be redistributed through the trajectory.

Mission planning can therefore influence SLAM quality.

Routes containing overlapping areas and return paths can provide stronger localisation than extremely long one-way trajectories.

Mapping Before Flight

Not every GNSS-denied system needs to build a map while flying.

Some environments can be mapped in advance.

A drone can then compare its observations with this reference map to determine its position.

This is sometimes called map-based localisation.

Warehouses, factories and other relatively stable environments are well suited to this approach.

The map can contain visual features, LiDAR geometry or other information.

However, substantial environmental changes may reduce matching performance.

Reference maps therefore need appropriate maintenance.

Terrain-relative navigation compares observed terrain with a known terrain model.

Sensors measure the landscape below or around the aircraft, and software searches for a corresponding pattern within stored geographic data.

This can provide an independent position reference.

The principle is particularly useful where terrain contains distinctive features.

However, flat or repetitive terrain provides fewer unique references.

The accuracy and currency of the reference map are also important.

Terrain-relative navigation should therefore be treated as one potential navigation source rather than a universal solution.

Visual Map Matching

Cameras can also compare observed features against previously mapped imagery.

Buildings, roads and other persistent features may provide geographic references.

Computer vision can determine where the observed scene best matches the stored map.

This potentially provides absolute rather than purely relative positioning.

However, appearance changes.

Lighting, seasons, construction, vegetation and weather can make the environment look different from the reference imagery.

Robust systems therefore use multiple features and confidence checks before accepting a match.

Radar can contribute to GNSS-denied navigation by measuring surrounding structures and terrain.

One major advantage is its ability to operate in darkness and through some difficult environmental conditions where optical cameras may perform poorly.

Radar may therefore complement cameras and LiDAR.

However, radar data has different characteristics from optical imagery and conventional LiDAR point clouds.

Interpretation and localisation require specialised processing.

As compact radar sensors continue improving, they are likely to become increasingly important for robust drone navigation.

Altimeters and Range Sensors

Knowing altitude or distance from surrounding surfaces can constrain the navigation solution.

Laser rangefinders, radar altimeters, ultrasonic sensors and barometric sensors can all contribute.

A downward-facing rangefinder can measure height above a floor or terrain.

A barometer estimates changes in altitude from atmospheric pressure.

These sensors do not normally provide complete three-dimensional positioning by themselves.

However, they reduce uncertainty in specific dimensions and therefore strengthen the overall sensor-fusion solution.

Magnetometers

Magnetometers measure the Earth’s magnetic field and are commonly used to estimate heading.

However, indoor and industrial environments can contain substantial magnetic interference from steel structures, electrical equipment and machinery.

This can make compass measurements unreliable.

GNSS-denied drones should therefore not depend exclusively on magnetic heading.

Visual, LiDAR or inertial systems may provide alternative orientation information.

The navigation system should detect inconsistent magnetometer data and reduce its influence when appropriate.

Ultra-Wideband Positioning

Ultra-Wideband, or UWB, can provide local positioning using installed reference devices.

Anchors are placed at known locations around an environment.

The drone measures signals from these anchors to estimate its position.

This can provide accurate indoor localisation.

Warehouses, factories, laboratories and controlled industrial environments are possible applications.

The disadvantage is infrastructure dependency.

UWB is not completely infrastructure-free because the anchors normally need to be installed and surveyed before operation.

However, for permanent facilities this may be entirely practical.

Radio-Based Local Positioning

Other radio technologies can also contribute to local positioning.

Known transmitters or network infrastructure can provide ranging or signal information.

However, signal strength alone is generally an unreliable measure of precise distance because walls, reflections and interference affect radio propagation.

Professional systems therefore use carefully designed ranging or positioning techniques.

Radio navigation can complement onboard perception but should be validated in the intended environment.

Indoor environments are among the most important applications for GNSS-denied navigation.

Warehouses, factories, aircraft hangars and large commercial buildings may contain no reliable satellite coverage.

A drone can combine VIO, LiDAR, optical flow and range sensing to navigate.

Pre-existing maps can improve localisation for repeat missions.

This creates opportunities for automated inventory, inspection and security applications.

However, indoor environments contain cables, pipes, cranes and other obstacles.

Navigation and collision avoidance therefore need to work together.

Warehouses can be challenging because aisles and racks may look highly repetitive.

Visual systems may see similar patterns repeatedly.

LiDAR may also observe similar geometric structures across multiple aisles.

Incorrect localisation can occur if the system mistakes one area for another.

Combining cameras, LiDAR, IMU and known warehouse maps can improve robustness.

Fixed reference systems such as UWB may also provide additional position information.

This makes warehouses strong candidates for multi-sensor navigation.

Industrial Facilities

Industrial plants contain pipes, machinery, vessels and complex structures that can provide rich visual and geometric features.

This makes them potentially well suited to SLAM navigation.

However, industrial facilities also contain reflective surfaces, moving machinery, poor lighting and electromagnetic interference.

No single sensor should therefore be assumed to work everywhere.

A drone may rely more heavily on cameras in one area and LiDAR in another.

Adaptive sensor fusion becomes particularly valuable.

Tunnels

Tunnels are a classic GNSS-denied environment.

LiDAR-inertial navigation is particularly useful because the walls provide continuous geometric measurements.

However, long tunnels can become geometrically repetitive.

Drift may accumulate along the tunnel axis.

Cross passages, equipment, changes in geometry and loop closures provide useful additional constraints.

For professional mapping, surveyed reference points may also be used to correct accumulated position error.

Underground Mines

Underground mines present one of the strongest commercial applications for GNSS-denied drones.

Drones can inspect stopes, tunnels and inaccessible voids without relying on satellite navigation.

LiDAR SLAM is widely suited to this environment because it works in darkness.

Cameras can provide additional visual information when suitable lighting is available.

However, mines introduce dust, water, uneven geometry and communications challenges.

Navigation therefore needs to be integrated with robust aircraft design and appropriate failsafe behaviour.

Confined Spaces

Tanks, vessels, utility chambers and other confined spaces can block GNSS completely.

A drone may use LiDAR, VIO and inertial navigation to move within them.

Protective cages are sometimes used to reduce damage from minor contact with surfaces.

However, navigation confidence may deteriorate inside geometrically uniform tanks.

Smooth cylindrical surfaces can provide fewer unique references.

The aircraft should therefore have conservative behaviour when localisation confidence decreases.

Under Bridges

GNSS reception can degrade beneath bridges and large infrastructure.

A drone inspecting the underside of a bridge may transition from good satellite positioning into partial or complete GNSS loss.

Hybrid navigation allows other sensors to maintain localisation during this transition.

Once the aircraft returns to open sky, GNSS can again contribute strongly.

This seamless transition between positioning sources is one of the most important capabilities for future infrastructure inspection drones.

Dense Urban Environments

Urban canyons can produce GNSS multipath, where satellite signals reflect from buildings before reaching the receiver.

The drone may still report a position, but that position can contain significant error.

This can be more challenging than complete GNSS loss because the navigation system needs to recognise unreliable information.

Visual and LiDAR navigation can provide independent observations.

A robust system should therefore evaluate GNSS quality rather than automatically trusting every available satellite solution.

Forests and Dense Vegetation

Tree canopy can reduce satellite visibility and create fluctuating GNSS performance.

Visual-inertial or LiDAR-inertial navigation can help maintain local positioning beneath vegetation.

However, forests create their own challenges.

Branches and leaves move in wind, while repetitive tree trunks may reduce visual uniqueness.

The system should therefore combine multiple navigation cues.

GNSS can continue contributing whenever signal quality improves.

Infrastructure Inspection

GNSS-denied navigation can expand inspection into areas that conventional waypoint drones struggle to reach.

Examples include beneath bridges, inside buildings, under roofs and around dense industrial infrastructure.

Once localisation is reliable, the drone can associate inspection observations with positions within a local 3D map.

This is particularly valuable when combining navigation with thermal, RGB, ultrasonic, gas or other inspection sensors.

The navigation system provides the spatial framework for the inspection data.

Construction Sites

Buildings increasingly become GNSS-denied as construction progresses.

A drone that initially maps an open site using GNSS may later need visual or LiDAR navigation once roofs and walls are installed.

Hybrid navigation allows the same platform to move between these environments.

The drone can compare captured geometry with BIM or design models.

However, construction sites change frequently.

Map-based localisation therefore needs to tolerate environmental changes.

Emergency Response

GNSS-independent drones can support emergency teams inside damaged buildings or infrastructure.

LiDAR can create a map while cameras and thermal sensors provide situational information.

The aircraft may explore areas considered unsafe for immediate human entry.

However, damaged structures can change while the drone is operating.

Dust, smoke and moving debris may also degrade sensors.

The drone should support responder awareness rather than be treated as confirmation that an area is structurally safe.

Search and Rescue

Indoor, underground and complex-terrain search operations can benefit from GNSS-denied navigation.

A drone can map its route while searching with RGB or thermal cameras.

The map can help responders understand where the aircraft has already searched.

However, detection of a thermal or visual feature does not automatically confirm a person’s identity or condition.

Navigation and sensing provide information for trained search-and-rescue teams to interpret.

Visible cameras require light.

Indoor and underground drones may therefore use onboard illumination.

Infrared-sensitive cameras may provide additional capability.

However, lighting creates shadows and reflections that can affect visual algorithms.

LiDAR provides an important complementary technology because it actively measures geometry.

Combining visual and LiDAR navigation can therefore provide greater resilience across changing lighting conditions.

Smoke and dust can degrade cameras substantially.

LiDAR can also be affected because particles scatter laser energy.

Radar may provide useful complementary sensing under some difficult visibility conditions.

No navigation sensor should be described as universally unaffected by smoke or dust.

The correct approach depends on particle density, sensor wavelength and environment.

Multi-sensor systems provide the strongest opportunity for continued navigation.

Water and Reflective Surfaces

Water, glass and polished surfaces can create problems for both cameras and LiDAR.

Visual systems may encounter reflections or limited texture.

LiDAR may receive weak or misleading returns.

Indoor environments containing large glass walls can therefore be challenging.

Navigation systems should combine information from multiple directions and sensors.

A missing LiDAR return should never automatically be interpreted as free space.

Dynamic Environments

Most SLAM systems work best when much of the environment remains stationary.

Moving vehicles, people, machinery and doors can introduce temporary features.

Advanced algorithms attempt to identify and ignore dynamic objects.

However, a busy warehouse or factory may still be more difficult than the same environment when inactive.

Navigation software should prioritise persistent environmental features.

This is particularly important for long-term autonomous operations.

Position Drift

Relative navigation systems accumulate error.

Even very good visual-inertial or LiDAR-inertial systems can gradually drift from the true position.

The rate depends on sensor quality and environmental conditions.

Loop closure, map matching, known landmarks and external references can reduce this error.

The key question is therefore not whether a GNSS-denied system experiences drift, but how effectively it detects and corrects it.

Mission duration and distance should be considered when evaluating navigation performance.

Relative Versus Absolute Position

GNSS normally provides a position in a global coordinate system.

SLAM may instead know that the drone is five metres from a wall and twenty metres from its starting point.

This is relative localisation.

For many indoor missions, relative position is sufficient.

For surveying or GIS integration, however, the local map may need to be connected to global coordinates.

Known control points or GNSS observations collected outside the building can provide this connection.

The distinction between local and global accuracy should always be understood.

Future autonomous systems increasingly need to understand their own navigation uncertainty.

Rather than reporting only a position, the system can estimate confidence in that position.

If camera features disappear, visual confidence may decrease.

If LiDAR geometry becomes repetitive, localisation uncertainty may increase.

The aircraft can then respond conservatively by slowing, stopping, returning toward a known area or requesting operator intervention.

This concept is fundamental to reliable autonomy.

Obstacle Detection

Navigation and obstacle detection are closely related but not identical.

A SLAM system may use LiDAR to understand environmental geometry, while a separate sensor provides immediate collision avoidance.

Cameras, stereo vision, LiDAR, radar and ultrasonic sensors can all contribute.

A drone knowing approximately where it is does not guarantee that it has detected every nearby obstacle.

Thin cables and transparent objects are particularly challenging.

Navigation and collision avoidance should therefore be evaluated independently.

Path Planning

Once the drone knows its position and surrounding geometry, software can plan a safe path.

The system identifies obstacles and calculates a route through available space.

For inspection applications, the route may also need to maintain a specified distance from the asset.

Path planning becomes increasingly important as drones move from manually controlled missions toward autonomous operations.

However, safe path planning depends on the accuracy and completeness of the environmental map.

Autonomous Exploration

An autonomous drone can potentially enter an unknown environment and determine where to explore next.

The SLAM map identifies areas already observed and areas that remain unknown.

Software can then select new viewpoints.

This could be valuable for mines, industrial inspections and emergency response.

However, autonomous exploration needs strong safety limits.

Battery reserve, communication quality, navigation confidence and return route should all be considered before the aircraft moves deeper into an unknown environment.

Returning safely can be more challenging than entering an environment.

GNSS-based return-to-home may not work underground or indoors.

A GNSS-denied drone may instead use its SLAM map or recorded trajectory.

It can potentially retrace its route toward the starting location.

However, the environment may have changed during the mission.

Moving machinery, doors or debris can block the original path.

A robust system should therefore re-evaluate the environment rather than blindly replaying previous control commands.

GNSS-independent navigation does not necessarily mean communications-independent operation.

A drone may know where it is while still losing its control link.

Underground mines and industrial structures can significantly reduce radio range.

Mesh networks, repeaters and other communications infrastructure can help.

Autonomous capability may also allow the aircraft to continue a safe procedure during temporary link loss.

Navigation and communications should therefore be treated as separate but interconnected engineering challenges.

Onboard Processing

GNSS-denied navigation often requires substantial onboard computation.

Camera frames, LiDAR scans and IMU measurements need to be processed continuously.

This places demands on the drone’s computing hardware.

Modern embedded GPUs and AI processors make sophisticated algorithms increasingly practical on small aircraft.

Onboard processing is particularly important because navigation cannot always depend on a remote cloud service or continuous high-bandwidth connection.

The aircraft needs to make critical localisation decisions locally.

Edge AI

AI can support feature recognition, environmental understanding and sensor-quality assessment.

For example, a system may identify stable structural features and give them greater importance than moving objects.

AI can also help classify obstacles and determine suitable flight areas.

However, AI should complement the underlying geometric and inertial navigation rather than act as an unexplained source of position.

Professional systems should maintain measurable navigation confidence and appropriate validation.

GNSS Spoofing and Interference Detection

A multi-sensor navigation system can also provide resilience when satellite information becomes inconsistent with other onboard observations.

For example, if GNSS indicates movement that strongly conflicts with inertial, visual or LiDAR estimates, the navigation system can identify the discrepancy and reduce reliance on the questionable source.

This type of cross-checking improves navigation integrity.

The objective is not simply operating without GNSS but ensuring that no single navigation source is trusted blindly when other sensors indicate that it may be unreliable.

Sensor Redundancy

Navigation reliability can be improved through redundancy.

A drone might contain multiple IMUs or complementary cameras.

LiDAR may provide an independent geometric reference.

Radar could provide additional sensing in difficult visibility.

However, adding sensors also increases weight, power consumption, calibration requirements and software complexity.

Redundancy should therefore be designed intelligently.

Several sensors that all fail under the same environmental condition may provide less resilience than two genuinely complementary technologies.

Time Synchronisation

Sensor fusion requires precise timing.

The drone moves continuously.

If a camera image, LiDAR scan and IMU measurement are associated with slightly different moments, the navigation estimate can become distorted.

Hardware synchronisation and accurate timestamps are therefore important.

This is particularly critical during rapid aircraft rotation or acceleration.

A sophisticated sensor suite with poor synchronisation can perform worse than a simpler but well-integrated system.

Sensor Calibration

The physical relationship between cameras, LiDAR, IMUs and other sensors must be known accurately.

This is called extrinsic calibration.

Camera characteristics also require intrinsic calibration.

If sensors are physically moved or mounts flex, these relationships can change.

Integrated navigation payloads therefore require mechanically stable installation.

Calibration should be checked when the system is modified or experiences significant impact.

GNSS-denied navigation can be integrated with specialist payloads.

A LiDAR mapping sensor may simultaneously support localisation and create a survey point cloud.

Thermal cameras can associate temperature observations with the local map.

Gas sensors can record concentrations at estimated positions.

Radiation sensors can build spatial hazard maps.

The navigation system therefore becomes the foundation connecting many different types of drone measurements to their physical locations.

GNSS-Denied Surveying

Mapping without continuous GNSS is possible, but georeferencing requires additional consideration.

SLAM can create a locally consistent map.

Known survey points can then connect that map to a global coordinate system.

This is particularly useful for tunnels and underground mines.

However, local SLAM accuracy and global survey accuracy are different concepts.

Professional surveying should independently verify the final coordinates.

Digital Twins

GNSS-denied navigation enables drones to update digital twins inside buildings and industrial facilities.

The drone can localise itself against an existing 3D model or create a new SLAM map.

Inspection observations can then be linked to assets within the model.

Repeated missions could monitor change automatically.

However, the digital twin needs accurate registration.

A visually convincing 3D model should not automatically be assumed to be dimensionally or globally accurate without verification.

Drone-in-a-Box Operations

GNSS-denied navigation could expand Drone-in-a-Box operations indoors.

A permanently installed drone could autonomously inspect warehouses, factories or large infrastructure.

The environment may already have a detailed reference map.

The aircraft could launch, localise against that map and perform repeat inspections.

Changes could be identified automatically.

This has the potential to turn indoor drone inspection from an occasional activity into a scheduled monitoring system.

Multiple drones may eventually collaborate inside GNSS-denied environments.

Each aircraft can contribute observations to a shared map.

One drone’s recognised features may help another improve localisation.

This could accelerate inspection of large mines or industrial facilities.

However, shared mapping introduces challenges involving communications, coordinate alignment and collision avoidance.

The system needs to ensure that every aircraft is operating within a consistent spatial reference.

Human Oversight

GNSS-denied autonomy should not eliminate human supervision where operational risk requires it.

Operators need information about navigation confidence, communications, battery condition and sensor status.

A simple position marker may not provide enough information.

The control interface should help the operator understand whether the aircraft’s localisation is becoming uncertain.

This enables appropriate intervention before a navigation problem becomes a flight-safety problem.

Data Security

Navigation maps of industrial facilities, infrastructure, mines and buildings may contain sensitive information.

SLAM datasets can reveal internal layouts and asset locations.

Cybersecurity should therefore be considered when storing or transmitting these maps.

Access controls, encryption and secure processing may be appropriate.

Navigation systems themselves also need protection against unauthorised software or data modification.

Reliable autonomy depends on both physical sensors and trustworthy digital systems.

The correct navigation technology depends heavily on the environment.

A warehouse may favour VIO, LiDAR and fixed infrastructure. A mine may rely heavily on LiDAR-inertial SLAM. A drone transitioning between outdoor and indoor environments may require GNSS, visual, LiDAR and inertial fusion. Difficult visibility may justify radar.

Important considerations include position drift, environmental feature requirements, lighting dependence, LiDAR range, IMU performance, map requirements, processing capability, obstacle detection, communications, aircraft endurance and failsafe behaviour.

Rather than asking whether a drone can “fly without GPS,” users should ask how the aircraft localises, how quickly its position estimate can drift, what conditions degrade that localisation, how the system detects uncertainty and what happens when navigation confidence becomes insufficient.

Benefits and Limitations

GNSS-denied navigation significantly expands where drones can operate.

It enables applications inside buildings, warehouses, tunnels, mines, industrial facilities, confined spaces, forests and infrastructure environments where conventional GNSS-based flight may be unreliable.

It can also provide additional resilience when satellite positioning becomes temporarily unavailable or inconsistent.

However, there is no universal replacement for GNSS.

Inertial navigation drifts. Cameras require usable visual information. LiDAR requires measurable geometry. Radar has different resolution characteristics. Optical flow depends on visible surface movement. Map matching depends on accurate reference data.

The strongest systems therefore combine several independent navigation technologies.

A drone that continues flying without GNSS should not automatically be considered safely localised. What matters is the quality and confidence of the alternative position estimate.

GNSS-denied navigation is likely to become a standard capability across a growing range of professional drones.

Advances in computer vision, compact LiDAR, radar, inertial sensors and edge computing are making sophisticated sensor fusion possible on increasingly small aircraft.

Future drones are likely to move away from thinking of GNSS as either available or unavailable. Instead, navigation systems will continuously evaluate many sources and dynamically adjust how much they trust each one.

A drone could begin a mission outdoors using GNSS and inertial navigation, transition beneath a structure using visual odometry, enter a dark tunnel using LiDAR-inertial SLAM and later reconnect its local trajectory with global coordinates when satellite visibility returns.

AI may assist with recognising stable landmarks and identifying unreliable sensor observations. Pre-existing digital twins could provide additional reference information. Collaborative drones may share maps, while Drone-in-a-Box systems could perform repeat autonomous inspections inside facilities.

A future workflow could operate as:

mission planning → GNSS-supported outdoor navigation → automatic detection of GNSS degradation → transition to visual/LiDAR/inertial localisation → real-time SLAM and obstacle mapping → continuous navigation-confidence assessment → specialist sensor data linked to the local map → loop closure and trajectory optimisation → reconnection with known reference points or GNSS → georeferenced mission dataset → professional review and digital-twin update.

Conclusion

GNSS-denied navigation is one of the key technologies expanding drones beyond conventional outdoor operations.

By combining inertial navigation, visual odometry, optical flow, LiDAR, SLAM, radar, range sensing and map-based localisation, drones can continue estimating their position in environments where satellite navigation is unavailable or unreliable.

This opens important opportunities in mining, tunnels, warehouses, factories, construction, infrastructure inspection, confined spaces, emergency response and autonomous indoor operations.

However, GNSS-denied navigation should not be viewed as a single technology or a perfect replacement for satellite positioning. Every alternative navigation method has conditions in which its performance deteriorates.

The strongest approach is therefore multi-sensor fusion with continuous confidence monitoring. Cameras can constrain inertial drift, LiDAR can provide geometry when lighting is poor, radar can provide additional environmental sensing, known maps can provide absolute references and GNSS can be reincorporated whenever reliable satellite positioning becomes available.

As these technologies continue to mature, drones will increasingly be able to transition seamlessly between outdoor, indoor, underground and other complex environments. GNSS will remain an important navigation source, but it will increasingly become one component within a broader and more resilient navigation architecture rather than the single technology on which autonomous drone operations depend.