Path Loss Map

A path loss map discretizes a geographic area into a raster grid of pixels or bins, where each cell stores a single dB value representing the predicted path loss from a specific transmitter. The resolution, typically between 1 and 50 meters, determines the granularity of the prediction. This structure allows for rapid lookup and visualization of coverage holes, enabling algorithms to quickly assess signal strength at any coordinate without re-running a full propagation model.

Path loss maps can be generated using two fundamentally different approaches:

• Deterministic models like ray tracing use 3D geometry and the physics of reflection, diffraction, and scattering to calculate precise paths. These are highly accurate but computationally expensive.
• Empirical models like the Okumura-Hata or 3GPP TR 38.901 models use statistical fits to measurement data, parameterized by distance, frequency, and environment type (urban, rural). They are fast but lack site-specific detail. Hybrid approaches combine both for a balance of accuracy and speed.

Path loss is fundamentally frequency-dependent. A map is generated for a specific carrier frequency; higher frequencies (e.g., mmWave at 28 GHz) experience significantly greater free-space path loss and atmospheric absorption than sub-6 GHz bands. This means a single physical environment requires separate path loss maps for each operating band. The relationship follows the Friis transmission equation, where free-space loss increases with the square of the frequency.

A pure path loss map provides only the large-scale, distance-dependent attenuation. To model realistic signal variability, it is combined with a shadow fading map, which adds a spatially correlated, log-normally distributed random variable to each grid cell. This shadow fading component accounts for the large-scale obstructions (buildings, foliage) not captured in the base propagation model. The combined map provides the total large-scale channel gain required for system-level simulations.

• Static maps are pre-computed for a fixed environment and transmitter configuration, suitable for long-term network planning and coverage optimization.
• Dynamic maps are updated in near-real-time to reflect environmental changes, such as a moving vehicle blocking a signal path or a new building being constructed. AI-enhanced RAN systems use dynamic path loss maps as a key input to predictive beamforming and proactive handover algorithms, allowing the network to anticipate signal degradation before it occurs.

In modern O-RAN architectures, path loss maps serve as a critical environmental state input to Deep Reinforcement Learning (DRL) agents hosted on the RAN Intelligent Controller (RIC). The DRL agent uses the map to understand the spatial relationship between users and cells, enabling it to learn optimal policies for load balancing, mobility management, and energy-saving cell sleep modes. The map translates raw geometry into a machine-readable format for neural network processing.

The mathematical engine behind every path loss map. A propagation model is a set of formulas that predict the average signal attenuation between a transmitter and receiver as a function of distance, frequency, and environment type.

• Empirical models (e.g., Okumura-Hata, COST 231) use measurement-fitted curves and are computationally light.
• Semi-deterministic models combine ray optics with statistical corrections for improved accuracy.
• Deterministic models (e.g., ray tracing) solve Maxwell's equations or geometric optics for site-specific precision.

The choice of model directly determines the resolution and fidelity of the resulting path loss map.

A deterministic propagation modeling technique that generates high-fidelity path loss maps by simulating individual radio wave paths through a detailed 3D environment.

• Accounts for specular reflection off building surfaces, diffraction around edges, and diffuse scattering from rough materials.
• Requires a precise 3D environment reconstruction with material properties (permittivity, conductivity) assigned to surfaces.
• Computationally intensive but essential for mmWave and sub-THz planning where wavelength-scale interactions dominate.

Ray tracing produces the most accurate path loss maps for dense urban and indoor scenarios.

A companion spatial grid that overlays the path loss map to model large-scale signal variations caused by obstructions not captured by the primary propagation model.

• Represents log-normal shadowing with a standard deviation typically between 6-12 dB, depending on environment clutter.
• Must exhibit spatial correlation: points close together experience similar shadowing, decaying with distance via an exponential autocorrelation function.
• Generated using correlated random fields (e.g., Cholesky decomposition or sum-of-sinusoids methods) to ensure realistic continuity.

Together, the path loss map and shadow fading map form the complete large-scale channel gain at any location.

A critical quality attribute ensuring that channel parameters evolve smoothly and continuously as a receiver moves through the environment. Without it, a path loss map becomes unrealistic for mobility simulations.

• Prevents abrupt jumps in path loss, delay spread, and angle of arrival between adjacent grid points.
• Achieved in Geometry-Based Stochastic Channel Models (GSCMs) by placing virtual scatterers in a geometric layout and updating paths continuously.
• Essential for valid handover simulation and beam tracking algorithm testing, where discontinuous channel snapshots would produce misleading results.

A spatially consistent path loss map enables the digital twin to support realistic drive testing scenarios.

A simulation methodology that replaces physical field measurements by replaying user mobility models through a digital twin equipped with a path loss map and channel emulation.

• A traffic generator creates synthetic data flows while the simulated UE moves along predefined routes.
• The path loss map provides the geometry-based signal levels at each timestep, feeding into link-level and system-level simulations.
• Dramatically reduces the cost and time of network validation while enabling testing of rare edge cases and extreme loading conditions.

Virtual drive testing is the primary operational use case for high-resolution path loss maps in RAN planning.

The foundational geospatial input for generating site-specific path loss maps via ray tracing. This process creates a digital three-dimensional model from raw sensor data.

• LiDAR point clouds from aerial or terrestrial scanning provide high-precision geometry.
• Photogrammetry extracts 3D structure from overlapping 2D images.
• GIS data (building footprints, heights, land use classifications) offers city-scale coverage at lower resolution.
• Material classification assigns electromagnetic properties to surfaces: concrete, glass, foliage, and metal each have distinct reflection and transmission coefficients.

The accuracy of the path loss map is fundamentally bounded by the fidelity of this reconstruction.