Joint BS Deployment and Power Optimization for Minimum EMF Exposure with RL in Real-World Based Urban Scenario
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Electrical Engineering. The day in electrical engineering: signals, systems and the hardware that carries them.
Rosa: We're looking at the paper: "Joint BS Deployment and Power Optimization for Minimum EMF Exposure with RL in Real-World Based Urban Scenario".
Paper summary: Rosa: So, we're looking at this paper titled "Joint BS Deployment and Power Optimization for Minimum EMF Exposure with RL in Real-World Based Urban Scenario." Basically, they're tackling the problem that as wireless systems get more complex and data rates go up, the electromagnetic field exposure from base stations is often underestimated, which has health implications.
Dev: Right. And this paper proposes a workflow that adjusts both where we put the base stations and how much power we use to minimize that EMF exposure while still maintaining good coverage. It's a joint optimization problem in a three dee city setting <ref:2504.05017#pg1>.
Rosa: The core idea is to make sure the deployment and power levels work together to hit those limits, which is important because you can't just optimize them separately without ruining one or the other.
Dev: They claim they achieved this by developing a novel least-time shoot-and-bounce ray, or SBR, ray launching tool. This tool helps with two things: it makes the computation much more efficient and it improves how they model diffraction for calculating EMF exposure accurately.
Taro: So you're using a shooting and bouncing method to model the paths? What does that mean for the complexity of figuring out where the energy actually goes in a complex urban environment?
Rosa: It means they use this SBR tool to get a better handle on those multipath effects, which is what causes diffraction—that's when waves bend around building edges or rooftops.
Dev: And to make sure they don't spend forever calculating every possible path, they also designed an adaptive grid refinement algorithm. This algorithm adjusts the resolution of the calculation based on how stable the effective channel is while considering how those base stations are beamforming.
Taro: So it's not just one big calculation, but a system that intelligently decides where to focus its attention based on what looks stable in the channel? That sounds like it could handle real-world variability better than a fixed mesh would.
Rosa: Exactly. And they pair all this up with a comprehensive and realistic EMF exposure model that actually takes into account things like the real transmit power, the interference between cells, and even how imperfect the channel state information is in practice.
Dev: That realism is key because if you use a simple model, you're just optimizing for a flawed picture. They then formulate this whole coverage-guaranteed EMF exposure minimization problem in a way that's accurate for deterministic channel models and solve it using an algorithm aware of the geometry.
Taro: So they are taking the physics modeling, which is pretty detailed—the line-of-sight, reflection, and diffraction stuff—and feeding it into this optimization framework to find a deployment and power setting that satisfies both the coverage need and the exposure constraint.
Rosa: That's the gist of it: improving efficiency with SBR for better physics modeling, using adaptive grids for speed over a large area, all wrapped up in a realistic model to solve for optimal deployment and power simultaneously.
Dev: It’s about making sure that when you design a network, you’re not just looking at signal strength or coverage, but also the actual physical impact of the electromagnetic field on people nearby. This paper is important because it moves the discussion from theoretical models to something that actually considers real-world physics and practical constraints.
Conclusion: Rosa: Looking back at this paper, "Joint BS Deployment and Power Optimization for Minimum EMF Exposure with RL in Real-World Based Urban Scenario," the authors are focused on how you can design a network that optimizes both signal quality and minimizes potential health risks from EMF exposure.
Dev: They did this by proposing a workflow that uses reinforcement learning to adjust base station placement and radiated power in a three dee urban setting to jointly satisfy both coverage and EMF limits <ref:2504.05017#pg1,and radiated power in a 3D urban>.
Rosa: The authors developed several key components for this, including the least-time shoot-and-bounce ray algorithm for faster and more accurate diffraction modeling, and the adaptive grid refinement algorithm for efficiently estimating exposure across the whole area.
Taro: So they are showing that you can use ray tracing techniques to get a better picture of where those fields are actually going to be, which is usually really hard in complex three dee spaces <ref:2504.05017#pg1>.
Dev: They also included a realistic model that accounts for things like actual transmit power and interference, and then they solved the resulting minimization problem using a geometry-aware algorithm adapted to deterministic channel models.
Rosa: The implication is that this approach offers a way to find deployment and power configurations that are physically sound regarding EMF exposure, not just mathematically clean.
Dev: It’s about making sure the system you build doesn't just connect devices well, but it also operates within safe physical constraints when people are around.
Taro: For someone listening who isn't a researcher, this means that when we plan where to put cell towers or small cells in a city, we can use methods that account for the actual physics of how radio waves bounce and bend off buildings, rather than just using simplified assumptions.
Rosa: It moves the discussion closer to practical deployment decisions by tying signal performance directly to physical exposure limits. This paper is about integrating these two very different engineering concerns—coverage and health—into a single optimization framework.
eess.SP
Submitted: 2025-04-07
Updated: 2026-11-17
License: http://creativecommons.org/licenses/by-nc-nd/4.0/
Importance score: 78/100
The gist: Base station (BS) deployment remains a critical task with successive wireless communication generations and increasing data rates demands…
Key concepts
- Least-Time Shoot and Bouncing Ray (SBR) Algorithm
- This is a novel reinforcement learning tool used to find the paths (multipaths) between a transmitter and receiver. It works by simulating rays that shoot out from the source and bounce off surfaces in the 3D environment, prioritizing paths that arrive at the destination in the shortest time. This helps accurately model how radio waves travel through complex urban structures.
- Adaptive Grid Refinement (AGR) Algorithm
- This algorithm intelligently adjusts the resolution of a computational grid used for calculations. It focuses on areas where signal coverage or EMF exposure is most unstable, refining the grid locally to get more precise results without wasting computing power everywhere. This improves accuracy in estimating signal strength and exposure across the entire urban area.
- EMF Exposure Model
- This is a detailed mathematical model that simulates how much electromagnetic field a person experiences. It goes beyond simple models by including realistic factors like actual transmit power, interference from other cells, and imperfections in channel information. This makes the simulation more representative of real-world network conditions.
- Ray-Based System Modeling
- This is the core physics model used to describe how radio waves propagate in a city with many obstacles. It accounts for all types of wave propagation, including direct line-of-sight transmission, reflections off buildings, and diffraction around corners. The field strength at any point is calculated by summing up the contributions from all these different paths.
Terminology
Summary
Base station (BS) deployment remains a critical task with successive wireless communication generations and increasing data rates demands, while the electromagnetic field (EMF) exposure is often underrated, yielding potential health implications This paper proposes a workflow that adjusts BS deployment and radiated power in a 3D urban scenario to jointly consider EMF exposure and coverage
Workflow Development
The proposed workflow involves several novel components designed to improve computational efficiency and accuracy for joint optimization. The major contributions include:
-
A novel RL simulation tool, incorporating a least-time shoot and bouncing ray (SBR) algorithm, is developed to improve computational efficiency and enhance diffraction modeling for accurate EMF exposure calculation
-
The adaptive grid refinement (AGR) algorithm is designed to efficiently estimate the EMF exposure and signal coverage across the urban area by adjusting the grid resolution based on the spatial stability of the effective channel while accounting for BS beamforming
-
A comprehensive and realistic EMF exposure model is presented that aligns with practical network behaviors, incorporating actual transmit power, intercell interference, and channel state information (CSI) imperfection
-
The coverage-guaranteed EMF exposure minimization problem is formulated in a realistic and accurate manner and solved using a geometry-aware algorithm adapted to deterministic channel models
Ray-Based System Modeling
The ray-based system model in the multipath scenario details the propagation phenomena, including Line-of-Sight (LoS) transmission, Reflection, and Diffraction. The E-field strength at the receiver (Rx) position is calculated by summing the contributions from all multipath components
((3) where ⃗ex, ⃗ey, ⃗ez are the unit vectors along x, y, z directions)
Ray-Launching Algorithm
The least-time SBR RL algorithm is developed to identify multipaths between Tx and Rx in a scenario similar to Fig. 1. Key aspects of the algorithm include:
- The geometrical scenario must be imported, defined by each triangle mesh vertices V
Real-World Validation
The RL tool is validated through real-world measurements, showing significant improvement when diffraction is incorporated. Specifically, the root mean squared error (RMSE) between the simulated and measured results is reduced from 6. p. 4 Previous simulators at millimeter
wave frequency omitted diffraction [18], given its reduced
significance at those frequencies. In our simulation, when
diffraction is not accounted for, the root mean squared
error (RMSE) between the simulated and measured results
is 6.Where the paper says it
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Adaptive Grid Refinement (AGR) for Global Estimation: The AGR algorithm is designed
based on the spatial stability of the effective channel while accounting for BS beamforming, enabling global estimation of EMF exposure and signal coverage.
This allows AI systems to efficiently estimate EMF exposure across a large urban area by dynamically adjusting grid resolution based on channel stability, reducing computational effort while preserving high estimation accuracy. -
Realistic Network Behavior Modeling: The system can incorporate
actual maximum transmit power
using thepower reduction factor (PRF)
derived from IEC 62232:2017 to obtain a statistical estimate of time-averaged BS transmit power, avoiding overestimation of time-averaged EMF exposure. -
Comprehensive Interference and Channel Imperfection Modeling: The AI system can accurately evaluate interference by deriving leakage formulas that account for
intercell interference
and modeling the degradation caused byCSI imperfection
using a linear SINR reduction factor based on estimation variance. -
Dynamic UE Mobility Integration: The system can capture realistic UE dynamics by employing a
correlated random walk (CRW) with von Mises distributed heading increments,
allowing the AI to model pedestrian behaviors like wandering and window shopping while incorporating their mobility over anEMF exposure averaging interval.
-
Coverage-Guaranteed Optimization via Nelder-Mead: The system can solve the coverage-guaranteed EMF exposure minimization problem by reformulating the non-convex constraint into a convex form with a penalty factor, then using the Nelder-Mead method to find optimal BS deployment and power configuration.
Abstract
Base station (BS) deployment remains a critical task with successive wireless communication generations and increasing data rates demands, while the electromagnetic field (EMF) exposure is often underrated. Therefore, this paper proposes a workflow that adjusts BS deployment and radiated power in a 3D urban scenario to jointly consider EMF exposure and coverage. To achieve this ambition, firstly, a novel least-time shoot-and-bounce ray (SBR) ray-launching (RL) tool is developed to improve computational efficiency, and simultaneously enhance diffraction modeling for accurate EMF exposure calculation, validated with real-world measurements. To efficiently extend the computation across the target urban area, the adaptive grid refinement (AGR) algorithm is designed based on the spatial stability of the effective channel while accounting for BS beamforming. Subsequently, to provide a more comprehensive evaluation of EMF exposure under practical network operations, a system-level abstraction further accounts for the actual maximum transmit power, intercell interference, and channel state information imperfection on the BS side, while capturing mobility over the EMF exposure averaging interval on the user equipment side. Upon the aforementioned aspects, the coverage-guaranteed EMF exposure minimization problem is formulated in a realistic and accurate manner, and solved by a geometry-aware algorithm that enables joint BS deployment and power optimization based on deterministic channel simulations under practical deployment constraints, completing the proposed workflow. Simulation results illustrate the application of the workflow in the considered urban scenario, while comparison with an empirical baseline highlights that the workflow mitigates EMF exposure underestimation and yields BS deployment configurations better aligned with real-world conditions.
Sources
- Wireless Hallucination in Generative AI-enabled Communications: Concepts, Issues, and Solutions
- An enhanced POSTA based on Nelder-Mead simplex search and quadratic interpolation
Related papers
- Open-H-RF: A Large-Scale Ultrasound Raw Channel Dataset Initiative
- Scalable Attention for 5G NR Channel Estimation
- Low-Altitude Wireless Networks: A Comprehensive Survey
- Sensing-Assisted OFDM with Index Modulation: A Waveform for Optical Wireless Integrated Communication and Localisation
- Iteration-Free Candidate Pruning and an Index-Domain VLSI Architecture for Sort-Free MIMO Box Decoding up to 1024-QAM
- A Foundation Model for Large-Scale Wireless Network Planning, Operation and Optimization