Source linked

Bayesian Optimization Puts 3 Base Stations on a Farm to Hit 72.8% Seasonal Coverage

arxiv.org@systems_wire3 hours ago·Machine Learning·0 comments

A GP-boosted placement algorithm achieves 72.8% worst-case coverage across four crop stages using only 50 ray-tracing evaluations, beating budget-matched approaches by at least 4.6 percentage points.

bayesian optimizationprivate 5g nragricultural iotray tracingprecision agriculturegaussian process

72.8% worst-case coverage across four seasonal stages with just three base stations, and it takes fewer than fifty ray-tracing simulations to find that placement. That's the result from a new paper tackling the problem nobody building private 5G NR for agriculture wants to admit: vegetation changes everything as the season progresses.

Vegetation Wrecks RF, But Nobody Bakes It Into Placement

Precision-agriculture networks rely on IoT sensor nodes that must stay connected through the entire crop cycle. The problem is that a cornfield in June looks nothing like that same field in August. Standard base-station placement treats the environment as static, so coverage that works at planting can collapse at harvest.

The authors formulate this as a maximin seasonal coverage problem: place $K$ base stations to maximize the worst-case coverage fraction across all crop growth stages. Each objective evaluation requires expensive ray-tracing simulations across every stage, making brute-force optimization intractable.

Gaussian Process Surrogate Cuts Simulation Cost by an Order of Magnitude

They adopt a Gaussian-process Bayesian optimization (GPBO) framework that builds a probabilistic surrogate of the robust objective using ray tracing results. This lets the algorithm explore candidate placements intelligently, focusing simulation budget where it matters.

On a $1,\text{km}^2$ multi-crop farm with three distinct crop zones at $3.5,\text{GHz}$, the scheme hits 72.8% worst-case coverage with $K{=}3$ BSs in fewer than fifty ray-tracing evaluations. That beats budget-matched state-of-the-art approaches by at least $4.6,\text{pp}$ across all four seasonal stages.

Private 5G for Agriculture Just Got a Practical Path to Deployment

The key insight is that you don't need a full EM simulation of every possible placement. A GP surrogate plus a few dozen ray-tracing runs gets you a deployment that survives the entire growing season. For any operator building private 5G NR on farmland, this turns an expensive multi-month planning exercise into something you can run on a laptop overnight.

Source: Robust Base Station Placement in Agricultural IoT via Bayesian Optimization
Domain: arxiv.org

Read original source ->

External source stays available while the OJO article and comment thread stay local.

More in Machine Learning

view topic
SNAP-FM Solves the Costly Constraint Problem in Physics-Constrained Generative Models

By exposing block-sparse Jacobian and KKT systems and solving them with ExaModels.jl and MadNLP.jl on GPUs, SNAP-FM accelerates nonlinear constraint projection without losing accuracy on PDE benchmarks.

Tuning the Classifier Alone Recovers 86% of Semi-Supervised Gains on Security Data

Bayesian optimization of the classifier and threshold achieves statistically equivalent results to full SSL pipeline tuning on 4 of 5 tabular security datasets.

Database Tricks Cut LLM Pipeline Costs by 85% With Proxy Models and AQP

Treating AI workflows as declarative queries lets approximate query processing and cheap decision trees reduce expensive model invocations by up to 90% on structured tasks, with less than 5% accuracy loss.

4,300 Coding Agent Sessions Expose Where LLM Serving Falls Short

University of Washington researchers release TraceLab: 350,000 LLM steps and 430,000 tool calls from real Claude Code and Codex usage, revealing long autonomous loops, short outputs, and heavy-tailed tool calls that...

Comments load interactively on the live page.