Localization in GNSS-Denied Environments

44 concepts47 papers1 proposal

Global Navigation Satellite Systems (GNSS), such as GPS, are the bedrock of modern outdoor navigation. However, their reliance on high-frequency, low-power radio signals creates a fundamental vulnerability: the signals cannot penetrate dense physical barriers. When a receiver moves indoors, underground, underwater, or into deep urban canyons, the line-of-sight connection to satellites is severed.

To maintain precise positioning without an orbital anchor, researchers must bypass satellite infrastructure entirely. This is achieved by exploiting local radio environments, tracking internal inertial movements, and matching sensory perception against structural maps.

The Physics of Signal Occlusion

To understand why alternative localization is necessary, we must look at how satellite signals interact with matter. GNSS signals operate at microwave frequencies that are easily absorbed or scattered by solid structures.

In terrestrial environments like tunnels, shopping malls, and urban canyons, concrete and steel completely block or reflect these waves. This reflection causes multipath interference, where a receiver catches the same signal multiple times at slight delays, destroying the time-of-flight calculations essential for GPS. As noted by Su et al. and Zhu et al., this degradation forces autonomous vehicles and mobile systems to rely on self-contained sensors.

Underwater, the problem is even more absolute. Saltwater is highly conductive, absorbing high-frequency electromagnetic waves within a few centimeters. As explained by Mirza et al., radio-frequency localization is fundamentally impossible sub-surface, requiring a shift toward acoustic or optical modalities.

Repurposing Local Infrastructure

If external satellite signals cannot penetrate a space, the next logical step is to utilize signals generated inside that space. Modern environments are filled with opportunistic radio frequency (RF) signals, such as WiFi, RFID, and Ultra-Wideband (UWB).

The most straightforward way to use these signals is through the Received Signal Strength Indicator (RSSI), which measures how much a signal attenuates over distance. By measuring RSSI from multiple known access points, a system can trilaterate its position, as demonstrated by Abdelhadi et al.. However, indoor spaces are dynamic; moving people and structural materials cause unpredictable signal fading. To overcome this noise, Sankhe et al. treat RSSI streams as time-series data, using Long Short-Term Memory (LSTM) networks to filter out multipath fluctuations and achieve sub-decimeter distance estimates.

[Dynamic Indoor Environment] 
       │
       ▼ (Multipath Fading & Shadowing)
[Noisy WiFi RSSI / CSI Signals] 
       │
       ▼ (Deep Learning / Sequence Models)
[Filtered, Precise Distance Estimates]

For higher precision, researchers look beyond signal strength to Channel State Information (CSI), which records how a signal scatters and reflects off specific obstacles. Gönültas et al. transform raw CSI data into spatial probability maps, bypassing hardware synchronization issues. To make these systems viable for real-time edge computing, Feng et al. utilize high-speed Kolmogorov-Arnold Networks (FasterKAN) to rapidly classify indoor locations with minimal latency. Alternatively, systems can use non-radio infrastructure: Pau et al. demonstrate that standard ceiling LEDs can be paired with fuzzy logic and particle swarm optimization to turn visible light into highly accurate positioning anchors.

Proprioceptive Tracking and Integration Drift

When no external infrastructure exists, a system must rely entirely on internal, proprioceptive sensors—such as Inertial Measurement Units (IMUs), wheel encoders, and barometers—to estimate its trajectory relative to a known starting point. This process is known as dead reckoning.

For example, Zhang et al. combine pedestrian step-detection algorithms with adaptive filtering to track smartphone users through dead reckoning. In deep subterranean environments like metro tunnels, where external signals are entirely absent, Ye et al. show that crowdsourced smartphone sensors can map unique magnetic and barometric anomalies to create a reliable fingerprint map for continuous tracking.

However, dead reckoning suffers from a fundamental mathematical limitation: integration drift. Because an IMU measures acceleration, calculating position requires integrating the signal twice over time. Any tiny, uncorrected sensor bias or random noise is integrated along with the true movement, causing the positioning error to grow quadratically. Without an external reference to bound this error, a dead-reckoning system will rapidly drift into invalid coordinates.

Exteroceptive Perception and Structural Priors

To bound integration drift, a system must observe its surroundings using exteroceptive sensors like LiDAR, cameras, or sonar. By mapping local geometric features and matching them against structural priors, the system can correct its accumulated dead-reckoning errors.

In terrestrial robotics, tightly coupling a 6 DoF IMU with LiDAR point clouds allows vehicles to navigate complex facilities with centimeter-level consistency, as shown by Frosi et al.. To ensure map-matching remains robust against sensor biases, Su et al. match scale-invariant curvature features of a vehicle's path against digital maps, focusing on the topological shape of the trajectory rather than absolute coordinate distances.

When explicit 3D maps are unavailable, architectural priors like 2D floor plans or Building Information Modeling (BIM) data serve as geometric constraints. Wang et al. demonstrate that a single monocular camera image can be matched against a 2D floor plan by jointly reasoning about store facades and text labels within a Markov Random Field. Similarly, Chiou leverages graph location networks to achieve zero-shot localization across unseen multi-view camera layouts by exploiting the underlying floor plan geometry.

[Inertial Sensors (IMU)] ───► [Dead Reckoning] ───┐
                                                   ▼
                                        [Factor Graph Optimization] ──► [Drift-Free Pose]
                                                   ▲
[Exteroceptive (LiDAR/Sonar)] ─► [Map Constraints] ┘

In underwater domains where optical sensors fail due to water turbidity, acoustic perception takes over. Cheng et al. convert dense multi-beam forward-looking sonar images into sparse point clouds to perform real-time Simultaneous Localization and Mapping (SLAM). To handle feature-sparse marine environments, Xu introduces an optimization framework that uses singular value decomposition to isolate well-constrained sonar keyframes, effectively bounding dead-reckoning drift even when visible landmarks are scarce.

Multi-Modal Sensor Fusion

The most robust localization frameworks do not rely on a single sensing modality. Instead, they mathematically fuse local RF infrastructure, proprioceptive tracking, and environmental perception into a unified state estimate.

The core challenge of multi-modal fusion is coordinate alignment. Different sensors operate in different coordinate spaces (e.g., a LiDAR's local frame versus a UWB beacon's global frame). Zou addresses this by using graph optimization to align local LiDAR point clouds with global UWB ranging coordinates via an IMU-assisted initialization phase.

By framing localization as a factor graph, different sensor observations can be treated as probabilistic constraints. For instance, Herath et al. fuse sparse WiFi location fixes, continuous smartphone IMU data, and floorplan images to reconstruct dense, highly accurate indoor movement histories. In distributed networks where centralized processing is impossible—such as underwater sensor arrays—zhang et al. utilize a factor graph framework to pass "weighted particles" between neighboring nodes, achieving cooperative positioning through local network consensus.

By binding local geometric observations, structural priors, and opportunistic radio signals into these unified probabilistic frameworks, modern localization systems successfully eliminate the need for an orbital satellite anchor.

Key Foundational Work

Indoor Positioning System Using Robust Outlier Extended Kalman Filter (2009) — contributed 1 concept elements to this direction.
Design, Development, and Testing of a Low Cost, Fully Autonomous Indoor Unmanned Aerial System (2010) — contributed 1 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_indoor_signal_attenuation_15f3d268,
  title = {indoor signal attenuation},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/indoor-signal-attenuation/rdr_15f3d2680f5b4f8f9cead92b609a3bc1},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-21},
}

This exact version (version 1, frozen text):

@misc{lacuna_indoor_signal_attenuation_15f3d268_v1,
  title = {indoor signal attenuation},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/indoor-signal-attenuation/rdr_15f3d2680f5b4f8f9cead92b609a3bc1/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-21},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake