Simulation Dataset for Scalability Boundaries in Dense EU868 LoRaWAN

This dataset supports the manuscript “Scalability Boundaries in Dense EU868 LoRaWAN: A Calibrated Airtime–Recovery Model with ns-3 Validation” (Mansour Abdulaziz, 2026). The repository contains the simulation results, provenance records, model outputs, and analysis artifacts used in the study.The experimental design comprises 2,020 ns-3.45 runs across 101 scenario cells, with 20 replications per cell. It evaluates uncoded single-channel operation, multi-channel reception, generation-based random linear network coding (RLNC), combined multi-channel RLNC, repetition controls, and confirmatory excess-coded RLNC cases. Simulation environment Simulator: ns-3.45 with the LoRaWAN module and study-specific measurement extensions Region / MAC: EU868, LoRaWAN Class A Deployment: static 500 m × 500 m field End-device positions: independently and uniformly distributed Gateways: three fixed gateways Propagation: log-distance model with exponent 3.76 and reference loss 7.7 dB at 1 m Additional shadowing random variable: none Spreading-factor assignment: static link budget Runtime adaptive data rate (ADR): disabled Uplinks: unconfirmed Application payload: 50 bytes Traffic periods: 10 s, 5 s, and 2 s Device counts: 50, 100, and 200 Run duration: 1200 s Measurement warm-up: first 100 s excluded Airtime accounting: conservative shared 1% per-device experimental uplink airtime budget across the evaluated channel set Operating profiles Baseline: single-channel uncoded uplink MC: uncoded multi-channel reception using eight evaluated EU868 channels NC: generation-based GF(256) RLNC on one channel MC–RLNC: generation-based GF(256) RLNC with multi-channel reception Rep: uncoded repetition control RLNC generation sizes are k∈{2,4,8}. The primary RLNC grid uses n=k, while confirmatory cases at k=4 evaluate excess scheduled coded transmissions with n>k.Each coded frame carries a 2-byte generation identifier and k GF(256) coefficient bytes, giving a coding-metadata overhead of k+2 bytes. Measurement semanticsLogical packet delivery ratio (PDR) is calculated from uniquely delivered or recovered original application packets. Multi-gateway receptions are deduplicated using the LoRaWAN device address and uplink frame counter.Physical-layer outcomes for unique uplink transmissions are reported separately from logical RLNC generation recovery. Queue-specific loss is not reported because the simulation traces do not support a validated queue-loss attribution. Duty-cycle limitation is therefore characterized through transmission admission and the fraction of generated traffic that does not become a unique on-air PHY transmission.Latency is conditional on successful logical delivery. Simulator latency values are converted from milliseconds to seconds for manuscript-facing analysis. Cells with zero logical PDR have no defined conditional delivery latency.Energy results are transmit-only (TX-only) metrics derived from transmit airtime. They do not represent total device energy, processing energy, receive energy, buffering energy, RLNC encoding/decoding energy, or complete battery lifetime. Repository contents CAMPAIGN_V5R_ALL_RUNS.csv — consolidated per-run results for all 2,020 runs CAMPAIGN_V5R_CELL_SUMMARY.csv — aggregate results for the evaluated scenario cells PLANNED_CAMPAIGN_V5R.csv — experimental design specification campaign_ledger.csv — execution and provenance information, including per-run SHA-256 fields MODEL_V6_1_FIT.csv — calibrated-model fitting results MODEL_V6_1_VALIDATION.csv — model-validation results MODEL_V6_1_RHO_VALIDATION.csv — coded-airtime-ratio validation results C_eff_sensitivity.csv — effective-parallelism sensitivity analysis model_mae_by_profile.csv — model-error summary by profile and generation size paired_MC_minus_baseline_PDR.csv — paired MC-versus-Baseline PDR analysis energy_per_packet_and_bit.csv — TX-energy-efficiency analysis model_v6_1.py — implementation of the calibrated airtime–recovery model analysis_v7.py — analysis script used to produce manuscript-facing results run_post_campaign_scientific_audit_v5r.py — scientific consistency and provenance audit CAMPAIGN_V5R_ALL_RUNS.csv provides the consolidated per-run evidence for the complete 2,020-run experimental design. Individual per-process CSV files are not included separately. Per-run provenance and SHA-256 information are retained in the consolidated campaign data and campaign_ledger.csv.The repository is intended to support reproducibility of the numerical analyses and conclusions reported in the associated manuscript.