Mitigating Positive Bias in Non Line of Sight Localization

115 concepts84 papers1 proposal

Modern indoor and urban localization systems—ranging from Ultra-Wideband (UWB) to 5G cellular networks—rely on a fundamental geometric assumption: radio signals travel in a straight line at the speed of light. In an ideal vacuum, measuring the Time of Arrival (TOA) of a signal allows a receiver to calculate its distance from a transmitter with surgical precision.

However, as Sinan Gezici notes in Position Estimation via Ultra-Wide-Band Signals, real-world environments are rarely ideal. When physical obstacles like concrete walls, metal machinery, or even human bodies block the direct path, the signal must reflect, diffract, or refract to reach its destination. This phenomenon, known as Non-Line-of-Sight (NLOS) propagation, introduces a systematic error that breaks the core logic of distance-based positioning.

The Geometry of the Positive Bias

The primary challenge of NLOS is not just noise, but positive bias. Because a reflected or diffracted signal must travel a longer, indirect path to circumvent an obstacle, the measured time of flight is always greater than the true straight-line distance. Hoon Choi et al. explain in Location Information Reliability-Based Precision Locating System Using NLOS Condition Estimation that this makes the source appear significantly further away than it actually is.

This bias is particularly problematic because it is non-Gaussian. While standard measurement noise can be averaged out, NLOS errors act as outliers that pull the estimated position toward the wrong coordinates. Michael Minyi Zhang et al. (Probabilistic Time of Arrival Localization) argue that traditional models fail because they do not account for the fact that these errors are inherently positive and environment-dependent.

Algorithmic Mitigation Strategies

To combat this, researchers have developed three primary families of solutions: robust optimization, adaptive filtering, and machine learning.

Robust Optimization

Traditional trilateration uses least-squares methods that are highly sensitive to outliers. To address this, Hongyang Chen et al. (Non-Line-of-Sight Node Localization Based on Semi-Definite Programming in Wireless Sensor Networks) propose using Semi-Definite Programming (SDP). By reframing localization as a convex optimization problem, their approach can utilize all available distance data—including corrupted NLOS measurements—without needing to identify them beforehand. Similarly, Chao Ding (Convex Euclidean distance embedding for collaborative position localization with NLOS mitigation) leverages Euclidean Distance Matrices to scale these robust solutions to networks with thousands of sensors.

Adaptive Filtering

In dynamic scenarios, such as tracking a moving robot or firefighter, filters must adapt to changing signal conditions. Sheng Xu et al. (A Double Extended Kalman Filter Algorithm for Weakening Non-Line-of-Sight Errors in Complex Indoor Environments Based on Ultra-Wideband Technology) introduce a Double Extended Kalman Filter (DEKF) that uses a residual classification module to adjust the filter's confidence in real-time. If a measurement suddenly jumps (indicating a new obstruction), the filter de-weights that specific beacon to prevent the position estimate from diverging.

Learning the Environment

Recent work has shifted toward "learning" the signature of NLOS signals. Wenzhong Yan et al. (Graph Neural Network for Large-Scale Network Localization) treat the wireless network as a graph, using Graph Neural Networks (GNNs) to regress node positions by learning which edges in the graph are likely corrupted by reflections. For energy-constrained devices, Zohreh Hajiakhondi-Meybodi et al. (DQLEL: Deep Q-Learning for Energy-Optimized LoS/NLoS UWB Node Selection) use Reinforcement Learning to train agents to ignore NLOS beacons entirely, saving battery life by only communicating with anchors that have a clear Line-of-Sight.

Multi-Modal and Structural Fusion

When radio signals alone are insufficient, researchers look to external constraints. Jinbo Hu et al. (Unscented-Kalman-Filter Based UWB and INS Fusion Positioning for Firefighters Application) demonstrate that fusing UWB with Inertial Navigation Systems (INS) allows the system to use the short-term reliability of motion sensors to "bridge" periods where UWB signals are heavily obstructed.

Furthermore, in multi-robot swarms, the relative geometry of the group can serve as a corrective force. Junho Choi et al. (SaWa-ML: Structure-Aware Pose Correction and Weight Adaptation-Based Robust Multi-Robot Localization) use inter-robot ranging to build a "rigid" geometric structure. This structure acts as a physical constraint: if one robot receives a biased NLOS measurement that suggests it has suddenly moved 5 meters away, the known distance to its peers can flag that measurement as an error.

The Path Forward

Despite these advances, extreme NLOS environments—where the direct path is completely extinguished—remain a significant hurdle. As Navid Tadayon et al. (Decimeter Ranging with Channel State Information) point out, without a detectable direct path, systems may erroneously lock onto reflected signals, leading to errors that even sophisticated filters struggle to correct. The future of the field likely lies in deeper integration of environmental mapping and multi-frequency analysis to turn these reflections from sources of error into sources of information.

Key Foundational Work

Position Estimation via Ultra-Wide-Band Signals (2009) — contributed 2 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_non_line_of_sight_signal_degradation_5377e3b5,
  title = {non-line-of-sight signal degradation},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/non-line-of-sight-signal-degradation/rdr_5377e3b50b194c319987790b6689bd5a},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-19},
}

This exact version (version 1, frozen text):

@misc{lacuna_non_line_of_sight_signal_degradation_5377e3b5_v1,
  title = {non-line-of-sight signal degradation},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/non-line-of-sight-signal-degradation/rdr_5377e3b50b194c319987790b6689bd5a/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-19},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake