PaperlayerView exact revision on arXiv

Source-linked AI summary

Battery-Aware Rate-Splitting Multiple Access for Solar-Powered Cell-Free LEO Satellite Networks

Wali Ullah Khan, Muhammad Adil

arXiv:2609.09445v1eess.SP

TL;DR

Solar-powered LEO downlinks face finite batteries and bounded harvesting uncertainty, while prior work does not explicitly couple one-layer RSMA power allocation to physical battery state. The paper develops fixed-cluster BA-RSMA with battery-aware control; in a 20-seed stressed experiment, it reduces near-depletion relative to Myopic-RSMA at a 1.7% EE penalty.

  • Problem

    Prior work does not explicitly couple one-layer RSMA common/private power allocation to finite satellite batteries under bounded harvesting uncertainty.

    Source: §I · p. 1

  • Method

    BA-RSMA couples common/private power allocation to the measured battery reserve in a fixed-cluster solar-powered cell-free LEO downlink.

    Source: §I · p. 1

  • Results

    BA-RSMA attains 1.631 Mbit/J and reduces mean near-depletion from 24.84% under Myopic-RSMA to 6.82%, at a 1.7% EE penalty.

    Source: C. Results Discussion · p. 5

  • Takeaways & Limitations

    Under the considered benchmarks, BA-RSMA exceeds NOMA and OMA in EE.

    Source: §V · p. 6

Abstract

from arXiv · show

Cell-free low Earth orbit (LEO) satellite downlinks improve coverage and macro-diversity, but intermittent solar harvesting and finite onboard batteries can make communication-only resource allocation energy-aggressive. We develop battery-aware one-layer rate-splitting multiple access (BA-RSMA) for a fixed-cluster solar-powered cell-free LEO downlink. A perturbed physical-battery Lyapunov queue couples common/private power allocation to stored energy, while robust energy causality protects against bounded harvesting uncertainty and allows harvest curtailment at battery saturation. Under a fixed-precoder scalar effective-channel model, dual and quadratic transforms yield convex fixed-auxiliary resource-allocation subproblems and monotonic alternating updates. Numerical cross-validation against a direct nonlinear-programming reference gives a maximum relative objective gap below $2.8\times10^{-7}$ over 12 representative slot--state instances. In a 20-seed battery-stressed experiment, BA-RSMA achieves 1.631~Mbit/J and reduces mean near-depletion from 24.84\% under Myopic-RSMA to 6.82\%, at only a 1.7\% EE penalty.

I. INTRODUCTION

Solar-powered LEO satellites have finite batteries, yet prior designs often use fixed transmit-power budgets and do not couple one-layer RSMA power allocation to physical battery state under bounded harvesting uncertainty. The paper proposes battery-aware RSMA for a fixed-cluster downlink and frames its scalar-channel model and resource-allocation approach.

I. INTRODUCTION

  • Prior designs do not explicitly couple one-layer RSMA common/private power allocation to finite satellite batteries under bounded harvesting uncertainty.This gap matters because energy-efficient transmission can remain aggressive before or during eclipse intervals.Source: §I · p. 1
  • BA-RSMA couples common/private power allocation to measured battery reserve in a fixed-cluster solar-powered cell-free LEO downlink.The formulation combines finite batteries, QoS constraints, and robust energy causality under bounded harvesting error.Source: §I · p. 1

A. Network and One-Layer RSMA Model

  • Fixed satellite association and beam directions reduce coherent transmission to scalar power allocation over an effective channel.The model treats desired and interfering streams at each UE as sharing an effective channel, a cluster-level power-control abstraction rather than general multiuser beamforming.Source: §I · p. 1, A. Network and One-Layer RSMA Model · p. 2
  • One-layer RSMA uses a common stream decoded by every UE alongside user-specific private streams, with the common rate limited by the weakest user.The common-message rate is split into nonnegative per-user portions.Source: A. Network and One-Layer RSMA Model · p. 2

B. Energy Model

The energy model assigns each satellite a share of aggregate radiated power and accounts for RF, fixed communication, and platform loads. It updates a finite physical battery with harvest curtailment at saturation and uses bounded harvest uncertainty to enforce robust energy causality.

B. Energy Model

  • Each satellite receives a fixed share of total radiated power, making RF communication load affine in aggregate stream powers.The share is computed offline from nominal service-region geometry and long-term statistics, and is a power-sharing proxy rather than an exact coherent-gain decomposition.Source: B. Energy Model · p. 2
  • The model excludes per-slot optimization of satellite power shares because it would reintroduce joint satellite-level power allocation beyond the scalar scope.The fixed proxy does not represent coherent-gain cross terms exactly.Source: B. Energy Model · p. 2
  • The physical battery accounts for harvested power, platform and communication loads, and curtails excess harvest when the battery is full.The battery state is capped at its maximum without requiring transmission solely to prevent overflow; stored energy uses a time base consistent with slot duration.Source: B. Energy Model · p. 2
  • Robust energy causality protects the battery reserve for every harvest realization within the stated error bound, provided the initial state and zero-radiated-power loads are feasible.The true battery update uses realized harvest, while the controller acts on a causal harvest estimate with bounded error.Source: B. Energy Model · pp. 2–3

A. Communication-Side Energy Efficiency

The paper measures communication-side energy efficiency using the communication loads in the battery model, while excluding the spacecraft bus load from the metric.

A. Communication-Side Energy Efficiency

B. Perturbed Battery Queue and Per-Slot Problem

A perturbed queue tracks the difference between the battery reserve target and stored energy, shaping a per-slot objective that balances communication-side efficiency with battery-aware energy consumption.

B. Perturbed Battery Queue and Per-Slot Problem

C. Fixed-Auxiliary Convex Reformulation

The paper transforms the coupled rate ratios and fractional EE objective into convex resource-allocation subproblems for fixed auxiliary variables, then alternates auxiliary updates with subproblem solutions.

C. Fixed-Auxiliary Convex Reformulation

A. Simulation Setup and Solver Evaluation

The simulation uses a three-satellite, four-user LEO cluster and evaluates the alternating solver through per-slot comparison with a direct nonlinear-programming reference.

A. Simulation Setup and Solver Evaluation

  • The simulated LEO scenario uses a three-satellite cluster serving four UEs over 96 slots, with a 20 GHz carrier and 20 MHz bandwidth.Source: A. Simulation Setup and Solver… · p. 4
  • Battery settings include a 35% eclipse fraction and bounded sunlit harvest uncertainty of 5 W; the battery-stressed comparison starts at 28 Wh with sunlit harvest of 40 W.Source: A. Simulation Setup and Solver… · p. 4
  • Fig. 2 plots average communication-side EE and minimum battery energy against V, alongside the corresponding EE–battery-reserve operating points.Source: Figure 2 · p. 4
  • For BA-RSMA long-horizon campaigns, curves are generated by the proposed dual-plus-quadratic-transform alternating algorithm, warm-started from the previous slot to reduce per-slot iterations, with relative outer stopping tolerance 10^-5.Source: A. Simulation Setup and Solver… · p. 4
  • The proposed algorithm's maximum relative objective gap to the direct reference was 2.8×10^-7 across 12 representative stressed slot–state instances.All 12 runs were monotonic and feasible, with a maximum final constraint residual of 9.3×10^-9.Source: A. Simulation Setup and Solver… · p. 4

B. Benchmark Models

The benchmark set separates the queue-aware battery term from other modeling choices and compares BA-RSMA with reduced-order RSMA, NOMA, and OMA implementations.

B. Benchmark Models

  • Continuous Myopic-RSMA uses the same feasible set with Zi(t) = 0 and a direct nonlinear-programming reference, isolating the queue-aware battery term.The remaining schemes are reduced-dimensional benchmarks.Source: B. Benchmark Models · pp. 4–5
  • Reduced-order BA-RSMA sets pc = βcP and splits the remaining power using βc and αp searched over specified finite grids.βc is searched over ten values in [0.05, 0.8], and αp ∈ {0, 0.5, 1, 1.5, 2, 3}.Source: B. Benchmark Models · p. 5
  • NOMA ranks users by channel gain, with each user cancelling weaker-ranked signals via SIC before decoding its own signal.Rank 1 is the strongest channel.Source: B. Benchmark Models · p. 5
  • OMA assigns equal resource shares τj = 1/U and computes each user’s rate using its allocated power and resource share.The benchmark schemes check each user's minimum-rate QoS and battery-dependent power cap at every candidate point; infeasible slots count as QoS outages.Source: B. Benchmark Models · p. 5

C. Results Discussion

In the battery-stressed experiment, BA-RSMA trades a small EE reduction against lower near-depletion than Myopic-RSMA. The section also reports the EE–reserve frontier, benchmark comparisons, and numerical validation of the alternating solver.

C. Results Discussion

  • As V rises from 0.1 to 100, average EE increases from 0.495 to 1.659 Mbit/J while minimum battery energy falls from 50.12 to 47.69 Wh.The EE–reserve frontier nearly flattens by V ≈100, and the sweep has no QoS outages or battery-floor violations.Source: C. Results Discussion · p. 5
  • 1.631 Mbit/J with 6.82% mean near-depletion contrasts with Myopic-RSMA’s 1.659 Mbit/J and 24.84%; BA-RSMA reduces near-depletion by 18.02 percentage points for about a 1.7% EE penalty.BA-RSMA's mean minimum battery state is 21.68 Wh versus 20.99 Wh for Myopic-RSMA; all schemes remain QoS-feasible in tested slots.Source: C. Results Discussion · p. 5, Figure 3 · p. 5
  • Reduced-order BA-RSMA achieves 1.437 Mbit/J and the lowest near-depletion rate, 4.01%, representing a more conservative RSMA operating point.Source: C. Results Discussion · p. 5
  • Under the considered benchmark implementations, BA-RSMA exceeds NOMA and OMA in EE by about 0.9% and 3.3%, respectively, with essentially the same near-depletion exposure.The ordering is not claimed as universal because NOMA and OMA use reduced-dimensional benchmark families.Source: C. Results Discussion · p. 5
  • Across 12 cross-validation instances, every alternating-solver run is monotonic and feasible, with a maximum relative objective gap of 2.8 × 10^-7 and maximum residual of 9.3 × 10^-9.Fig. 4 shows convergence for a representative stressed slot toward the direct SLSQP reference solution.Source: C. Results Discussion · pp. 5–6, Figure 4 · p. 5

V. CONCLUSION

  • The conclusion reports that battery-aware one-layer RSMA couples power allocation to stored energy and uses robust energy causality with harvest curtailment to maintain battery safety.It identifies dynamic clustering and geometry-based multilayer RSMA as future work.Source: §V · p. 6
Loading 2609.09445v1…