Pith. sign in

REVIEW 4 major objections 5 minor 19 references

Cross-Layer Optimization and System-Level Design of Next-Generation Wireless Networks via Intelligent RAN Control

T0 review · 4 major / 5 minor · reviewed 2026-08-04 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Deep-RL xApps trained offline can lift eMBB throughput by ~99% and mMTC by ~56% in emulated Open RAN, with gains tuned by reward weights and control timers.

desk verdict A solid compilation dissertation whose new increment is a large DRL-xApp benchmark and first RIS-in-Colosseum experiments; the headline gains are real inside the emulator but should not be read as field predictions until external fidelity is addressed. read the letter →

arxiv 2608.02067 v1 pith:2EQIN7AF submitted 2026-08-03 cs.NI

Maria Tsampazi This is my paper
classification cs.NI
keywords OpenRANdeepreinforcementlearningnetworkslicingschedulingxAppsIntelligentControllerdigitaltwinreconfigurablesurfaces
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This dissertation tries to establish that Open RAN near-real-time controllers can run deep-reinforcement-learning applications, trained offline on emulator-collected telemetry, to make slicing and scheduling decisions that yield large per-slice gains in a hardware-in-the-loop emulator. It argues that performance is not fixed by the learning algorithm alone: action space, discount factor, reward weights, and the timers governing RAN reporting and action enforcement are first-order design axes. The PandORA pipeline automates training and on-boarding of these agents as xApps. If true, this makes AI-native RAN control a design-space problem with concrete levers, and it supports the broader vision of software-defined, reconfigurable RANs for 6G.

What carries the argument

The carrying mechanism is PandORA, an automated pipeline that parses a JSON intent (slices, KPIs, actions, reward weights), trains DRL agents (PPO and DQN) offline on KPM datasets, packages them as xApps, and dispatches them into a near-real-time RIC controlling an srsRAN-based RAN emulated on Colosseum with FPGA-based channel emulation. The DRL agent's state is an autoencoded vector of three KPIs (DL throughput, buffer occupancy, transmitted packets); its action space spans slicing (PRB allocation among eMBB/mMTC/URLLC) and scheduling profile (RR, WF, PF); reward is a weighted sum with a negative URLLC weight to minimize buffer occupancy as a latency proxy. RAN control timers set the period

What would settle it

Run the same trained xApps (e.g., Sched & Slicing 0.5) in a real over-the-air small-cell testbed with comparable traffic and mobility, and compare achieved eMBB throughput and mMTC packet counts with the emulator's numbers; if the ~99%/~56% gains collapse or reverse, the central claim fails. A cheaper check: measure the CIR mismatch between MCHEM's 4-tap approximation and a ray-traced or measured channel in the same topology.

Watch

Extended reading notes

Core claim

The central claim is that DRL-based xApps, trained offline and deployed on the near-real-time RIC, can substantially and tunably improve network slicing and scheduling. In the author's experiments, finer decision-making granularities improve mMTC performance by ~56% and increase eMBB throughput by ~99% (Section 2.2.1), while URLLC buffer occupancy remains at zero. The paper further claims that these gains are systematically shaped by design choices—reward weights, action space, discount factor, DRL architecture (PPO vs DQN), hierarchical vs joint control, RAN control timers, and per-slice vs joint optimization—and that the PandORA framework automates the training and on-boarding of these age

Load-bearing premise

The results assume that the Colosseum emulator—with its at-most-4-tap channel model, emulated traffic, and aggregate KPI feedback—behaves faithfully like a real deployment; if emulated channels and reporting differ from the field, the relative gains will not transfer.

Editorial extensions

If this is right

  • RAN operators can treat DRL design choices as tunable levers: short-horizon rewards (gamma=0.5) plus joint slicing-and-scheduling favor eMBB throughput, while long-horizon slicing-only favors mMTC.
  • Fine-grained control timers (~100 ms) can roughly double scheduling-controlled eMBB throughput (~99% increase) without hurting mMTC, making control-loop cadence a first-order parameter.
  • The PandORA intent-driven pipeline implies that xApp development can be automated from a JSON spec, reducing the manual effort of training and deployment.
  • DRL-based xApps with shared optimization intents produce only low-severity conflicts (KPM distances below 0.5), suggesting that coexistence of multiple intelligent controllers is manageable.
  • Out-of-sample results (new location, mobility, traffic load) show robustness, with two hierarchical xApps outperforming a single joint xApp under mobility.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the gains are real, the biggest open question is transfer to field: the emulator's 4-tap TDL channels and aggregate-KPI observations may not capture real multipath and channel-state dynamics, so one would expect smaller or different gains over-the-air; a direct emulator-to-OTA comparison would settle this.
  • The finding that finer timers boost throughput suggests an implicit trade-off with RIC and E2 interface load; at scale, 100 ms control loops on many cells could strain the near-real-time RIC, so a cost-aware control-cadence scheduler is a natural next step.
  • The strong sensitivity to reward weights and timers implies an opportunity for meta-optimization—searching reward weights and control timers per deployment—which the paper's catalog-based approach already hints at.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The dissertation argues that AI-native Open RAN control, implemented as DRL-based xApps on a near-real-time RIC, can deliver substantial and tunable gains for network slicing and scheduling. Chapter 2 introduces PandORA, an automated pipeline for training and deploying DRL xApps, and benchmarks 23 xApp designs on the Colosseum emulator, reporting gains such as ~56% mMTC improvement and ~99% eMBB throughput increase when varying action spaces, reward weights, and RAN control timers. Chapter 3 extends the work to RIS-assisted channels, combining QuaDRiGa channel modeling, a Stackelberg game for energy-efficient power control, and Colosseum experiments. Chapter 4 proposes ARIADNE, an RL link-adaptation agent evaluated on ray-traced channels and OTA data. Chapter 5 studies FR3 spectrum sharing between terrestrial and NTN systems via interference nulling and QoS-aware power control. The central load-bearing claim is that DRL xApps, trained offline and deployed in the near-real-time RIC, provide robust, tunable performance gains for heterogeneous slices.

Significance. If the central claim holds, the work would provide a useful design methodology for AI-native RAN control: it systematically analyzes how reward weights, action spaces, discount factors, and RAN control timers affect slice-level performance, and it releases a framework (PandORA) that automates much of the xApp lifecycle. The experimental campaign is extensive (38+ hours, 23 xApps, in- and out-of-sample tests), and the use of Colosseum with hardware-in-the-loop is a strength relative to pure simulation. The RIS and spectrum-sharing chapters add breadth, with reproducible channel modeling and game-theoretic formulations. However, the headline gains are median point estimates from a single emulator, and the external validity of Colosseum's 4-tap TDL emulation to real deployments is not established; this limits the strength of the claimed transferability.

major comments (4)
  1. [§2.2.5, Figs. 2.5–2.16] The headline results (e.g., ~56% mMTC improvement, ~99% eMBB throughput increase) are reported as median point estimates, with no confidence intervals, variance measures, or significance tests. Given the stochastic nature of DRL and channel emulation, the claim that these differences are attributable to the design axes (action space, discount factor, timers) rather than noise is not statistically supported. Please provide per-configuration distributions (boxplots or CIs) and, where appropriate, a paired significance test across repetitions.
  2. [§2.1.5 and §2.1.3] The RF-twinning validation in §2.1.5 compares the emulated CIR and path loss against the originally modeled taps from the ray tracer, not against measurements from a deployed network. This verifies internal consistency, not external fidelity. Since the DRL agents observe only aggregate KPIs and channel state is only 'indirectly captured' (§2.2.2.3), the transfer of relative gains to real deployments is an untested assumption. Please add a direct comparison with field measurements, or explicitly re-scope the claims to emulator-relative improvements and discuss the conditions under which they would transfer.
  3. [§2.2.5, URLLC results (e.g., Figs. 2.5c, 2.7c, 2.15c)] The manuscript repeatedly reports a 'median buffer occupancy of 0 byte' across nearly all configurations and interprets this as 'optimal' URLLC performance. Because buffer occupancy cannot go below zero, the median is a floor effect and carries little information about tail latency or reliability. The reward in Eq. (2.6) uses buffer occupancy as a proxy, but no end-to-end latency or tail-latency metric is reported. Please report the full distribution or high percentiles of buffer occupancy, and ideally a latency metric, before claiming URLLC optimality.
  4. [§2.3, Table 2.14 and Fig. 2.29] The conflict analysis uses PACIFISTA—a framework developed by the same authors—to profile and characterize the authors' own xApps. While this is not circular in a mathematical sense, the reported 'low-severity' conflicts (distances below 0.5) lack calibration against an independent baseline or a threshold validated on real deployments. The analysis is limited to three xApps and two KPMs, so the general claim that DRL-based xApps with shared intents produce minimal conflicts is not yet supported. Please provide a baseline comparison (e.g., against random policies or non-DRL controllers) and justify the severity thresholds.
minor comments (5)
  1. [§2.2.4, Table 2.2] The Default weight for eMBB (72.0440333) and URLLC (0.00005) are reported with excessive precision and no derivation; the Alternative weight for mMTC is explained via Table 2.6, but the eMBB and URLLC values remain unexplained. A one-sentence derivation or citation would help.
  2. [§2.2.7, Table 2.12] The xApp Catalog table is dense and wraps awkwardly in the text; consider splitting it into multiple tables or using a landscape layout. Also, the Testing Conditions column mixes several dimensions (mobility, timers, location) without clear separators.
  3. [§3.1.3.2, Eqs. (3.4)–(3.7)] The utility function parameters α and M are described as controlling the slope, but no values are given in the text or Table 3.1. Since the game-theoretic results depend on these free parameters, please provide the values used and a sensitivity analysis.
  4. [§4.4, Table 4.2] The performance comparison 'with respect to the Oracle' would benefit from a definition of the oracle and the error bars for each entry; ARIADNE's spectral efficiency and BLER are otherwise only shown graphically in Fig. 4.3.
  5. [General] The manuscript contains many repeated acronym definitions and some figures are referenced before being introduced (e.g., Fig. 2.3 is discussed in §2.2.2 but appears later). A final proofreading pass for cross-reference consistency is recommended.

Circularity Check

0 steps flagged · score 0.0 of 10

No equation-level circularity: the DRL gains are measured emulator outcomes; self-citations are supporting rather than load-bearing, and the main validity caveats concern external fidelity, not circular reasoning.

full rationale

The dissertation's central claims are empirical results from Colosseum experiments, not derivations from first principles that collapse into their own inputs. The reported gains (e.g., ~99% eMBB throughput increase for Sched 0.99 from Set 2 to Set 3 in Section 2.2.5.5, and ~56% mMTC improvement from timer changes in Section 2.2.6) are direct comparisons of measured medians under different control timers, action spaces, or reward weights; no equation defines a predicted quantity in terms of a fitted quantity. The Alternative weight configuration (Section 2.2.5.4) uses historically maximum KPM values as normalizers, but this is reward scaling, and the subsequent performance difference is measured, not a fitted parameter renamed as a prediction. Self-citations to [26] and [2] supply the xApps and the PACIFISTA conflict framework, but the conflict distances in Table 2.14 are re-measured on Colosseum, not imported as conclusions; no uniqueness theorem or ansatz is smuggled in via citation. The genuinely weak points—the RF-twinning validation in Section 2.1.5 compares emulated CIR/path loss to the originally modeled taps, and the DRL state only indirectly captures channel state (Section 2.2.2.3)—are external-validity limitations, not circular reasoning: matching an emulator to its own input model checks internal consistency but does not validate the model against field measurements. Since the paper does not exhibit any reduction of a claimed result to its own inputs, the circularity score is 0.

Assumptions & free parameters 6 free parameters · 6 assumptions · 0 invented entities

The dissertation's central claims rest on four main classes of unpaid premises: fidelity of the Colosseum emulation chain (4-tap TDL channels), proxy metrics (buffer occupancy for latency, aggregate KPIs for channel state), the accuracy of the QuaDRiGa-based RIS channel cascade, and hand-chosen design parameters (reward weights, utility slopes, DRL hyperparameters) that directly shape the reported gains. None of these are machine-checked or independently validated outside the authors' own toolchain.

free parameters (6)
  • Default reward weights = eMBB 72.0440333, mMTC 0.229357798, URLLC 0.00005
    Hand-chosen weights (Table 2.2) that directly determine which xApp configuration appears 'best' in the benchmark.
  • Priority factors for Alternative weights = alpha_eMBB=1000, beta_mMTC=456, gamma_URLLC=1
    Hand-chosen scaling factors (Table 2.6) combined with historical maxima A, B, C from the same KPM datasets to produce the Alternative weights.
  • Historical maxima A, B, C = A=13.88 Mbps, B=304 packets, C=20186 bytes
    Fitted to the evaluation KPMs (Section 2.2.5.4); used to construct the Alternative weight configuration later reported as a design that improves mMTC.
  • Utility slope parameters alpha and M = Not given in text
    Appear in the energy-efficiency utility of Eq. (3.4) and control the sigmoid shape; values are not specified, so the reported game-theoretic power control results cannot be reproduced from the text.
  • DRL hyperparameters = learning rate 1e-3, gamma in {0.5, 0.99}, eps_DQN=0.1, replay buffer 10,000, networks 3x30 and 5x50
    Standard choices asserted without sensitivity analysis; results like the PPO-vs-DQN ranking depend on them.
  • RAN control timers = Set1: 1s/250ms/250ms; Set2: 250ms/250ms/250ms; Set3: 100ms/100ms/100ms
    Treated as the variable under study, but the three discrete settings are chosen by hand and conclusions depend on the sampled granularities (Table 2.9).
assumptions (6)
  • domain assumption Colosseum MCHEM emulation (tapped-delay-line channels with at most 4 taps) faithfully replicates real-world RF propagation
    All performance gains are measured through this emulator; validation in Section 2.1.5 compares emulated CIR and path loss to the modeled taps, not to measurements from a deployed network.
  • domain assumption Aggregate KPIs without explicit CSI are sufficient inputs for DRL control
    Section 2.2.2.3 asserts channel effects are 'indirectly captured' by throughput, buffer occupancy, and transmitted-packet KPIs, citing [11] without verification in this text.
  • domain assumption BS buffer occupancy is a valid proxy for URLLC latency
    Stated in the footnote of Section 2.2.2.2: the eNB lacks end-to-end latency, so buffer occupancy is used; a zero median buffer is then read as 'optimal' URLLC performance.
  • domain assumption QuaDRiGa 3GPP 38.901 UMa GBSM with coherent cascaded LOS links accurately models RIS-assisted channels
    Section 3.1.3.1; the frequency-dependent modeling is asserted accurate without measurement-based validation, and the use of omni-directional antennas at mmWave is acknowledged by the author as atypical.
  • standard math NOMA SIC decoding order and a Stackelberg equilibrium describe the UL power control game
    Section 3.1.3.2; standard model, but equilibrium existence and uniqueness are not proven in the text.
  • standard math PPO clipped surrogate objective and DQN Bellman updates are appropriate for the RAN control MDP
    Section 2.2.2.3; standard RL machinery assumed to converge to useful policies after offline training.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Cross-Layer Optimization and System-Level Design of Next-Generation Wireless Networks via Intelligent RAN Control." pith.science (2026). https://pith.science/paper/2EQIN7AF

@misc{pith2026260802067,
  author       = {Pith},
  title        = {Pith review of: Cross-Layer Optimization and System-Level Design of Next-Generation Wireless Networks via Intelligent RAN Control},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2EQIN7AF}},
  note         = {Machine review of arXiv:2608.02067}
}
read the original abstract

Recent years have seen the evolution of the traditional Radio Access Network (RAN) toward more open, programmable, disaggregated, and intelligent architectures, known as an Open RAN. Future Next Generation (NextG) networks are envisioned to be AI-native, enabling data-driven closed-loop optimization of Base Station resources, while Reconfigurable Intelligent Surfaces (RIS) emerge as key enablers for wireless propagation and spectral efficiency toward 6G and beyond. This dissertation focuses on the design, optimization, and experimental evaluation of NextG RANs integrating Open RAN principles, data-driven control loops, and intelligent resource allocation. The work emphasizes cross-layer optimization, including energy-efficient power control, and explores AI-driven network slicing, scheduling, and link adaptation, demonstrating NextG RANs reconfigurable in real time to meet 6G requirements, first analyzing architectural enablers and modeling frameworks, then prototyping and evaluating solutions on experimental platforms and Digital Twins. Main contributions include: (i) Deep Reinforcement Learning (DRL) solutions for network slicing and scheduling; (ii) PandORA, a framework for automatic design, training, and deployment of DRL-based Open RAN applications on the Colosseum wireless network emulator; (iii) physical-layer RIS channel modeling and optimized resource allocation across spectrum bands; (iv) system-level evaluation of RIS-assisted channels for eMBB and URLLC traffic; (v) integration of RIS within Open RAN; (vi) online RL solutions for link adaptation; and (vii) spectrum sharing between cellular and Non-Terrestrial Network links via power control and beamforming. This work provides algorithmic designs, frameworks, and validation from simulation and hardware-in-the-loop emulation to over-the-air 5G testbed experiments, addressing industry and academic needs for wireless research.

Figures

Figures reproduced from arXiv: 2608.02067 by the authors.

Figure 1.1
Figure 1.1. NextG Research Framework: Requirements, Mechanisms, Architecture, Enablers, and [PITH_FULL_IMAGE:figures/full_fig_p020_1_1.png] view at source ↗
Figure 1.2
Figure 1.2. The Open RAN Paradigm: programmable, virtualized, and disaggregated NextG cellular [PITH_FULL_IMAGE:figures/full_fig_p023_1_2.png] view at source ↗
Figure 2.1
Figure 2.1. Open RAN twinning capabilities in Colosseum. [PITH_FULL_IMAGE:figures/full_fig_p027_2_1.png] view at source ↗
Figures from the paper (76 more)
Figure 2.2
Figure 2.2. Figure 2.2: PandORA framework for intent-driven DRL training, xApp on-boarding, and testing [PITH_FULL_IMAGE:figures/full_fig_p037_2_2.png]
Figure 2.3
Figure 2.3. Figure 2.3: Reference O-RAN testing architecture with focus on the case of two xApps operating at [PITH_FULL_IMAGE:figures/full_fig_p046_2_3.png]
Figure 2.4
Figure 2.4. Figure 2.4: Reference O-RAN testing architecture with focus on the case of four xApps operating at [PITH_FULL_IMAGE:figures/full_fig_p047_2_4.png]
Figure 2.5
Figure 2.5. Figure 2.5: Performance evaluation under different action spaces and values of the [PITH_FULL_IMAGE:figures/full_fig_p052_2_5.png]
Figure 2.6
Figure 2.6. Figure 2.6: Median values under different action spaces and values of [PITH_FULL_IMAGE:figures/full_fig_p052_2_6.png]
Figure 2.7
Figure 2.7. Figure 2.7: Performance evaluation under different hierarchical configurations with the PPO DRL [PITH_FULL_IMAGE:figures/full_fig_p055_2_7.png]
Figure 2.8
Figure 2.8. Figure 2.8: Median values under different hierarchical configurations with the PPO DRL Architecture. [PITH_FULL_IMAGE:figures/full_fig_p055_2_8.png]
Figure 2
Figure 2. Figure 2: a. With respect to mMTC, Figs. 2.9b and 2.10b depict a similar performance to the one [PITH_FULL_IMAGE:figures/full_fig_p056_2.png]
Figure 2.9
Figure 2.9. Figure 2.9: Performance evaluation under the Default weight configuration for different action spaces and discount factor γ = 0.95 with the DQN DRL Architecture. (a) eMBB DL Throughput (b) mMTC Packets [PITH_FULL_IMAGE:figures/full_fig_p057_2_9.png]
Figure 2.10
Figure 2.10. Figure 2.10: Median values under the Default weight configuration for different action spaces and discount factor γ = 0.95 with the DQN DRL Architecture. 38 [PITH_FULL_IMAGE:figures/full_fig_p057_2_10.png]
Figure 2.11
Figure 2.11. Figure 2.11: Performance evaluation with 4 xApps and per-slice scheduling profile selection under PPO and DQN Architectures. (a) eMBB Throughput (b) mMTC Packets [PITH_FULL_IMAGE:figures/full_fig_p058_2_11.png]
Figure 2.12
Figure 2.12. Figure 2.12: Median values obtained with a 4-xApp setup and per-slice scheduling profile selection under PPO and DQN Architectures. 39 [PITH_FULL_IMAGE:figures/full_fig_p058_2_12.png]
Figure 2.13
Figure 2.13. Figure 2.13: Performance evaluation under the Alternative weight configuration for different action spaces and discount factors with the PPO DRL Architecture. In this study, we consider different weight configurations to compute the cumulative average reward function in Eq. (2.6…
Figure 2.14
Figure 2.14. Figure 2.14: Median values under the Alternative weight configuration for different action spaces and discount factors with the PPO DRL Architecture. so as to scale the weight according to the dynamic range of the corresponding KPM. Similarly, αeMBB, βmMT C, and γURLLC can be us…
Figure 2.15
Figure 2.15. Figure 2.15: Performance evaluation under different action spaces, values of the [PITH_FULL_IMAGE:figures/full_fig_p064_2_15.png]
Figure 2.16
Figure 2.16. Figure 2.16: Median values under different action spaces and RAN control timers with the PPO DRL [PITH_FULL_IMAGE:figures/full_fig_p065_2_16.png]
Figure 2.17
Figure 2.17. Figure 2.17: Performance evaluation under mobility with Settings [PITH_FULL_IMAGE:figures/full_fig_p067_2_17.png]
Figure 2.18
Figure 2.18. Figure 2.18: Median values obtained under mobility with Settings [PITH_FULL_IMAGE:figures/full_fig_p067_2_18.png]
Figure 2.19
Figure 2.19. Figure 2.19: Performance evaluation focusing on the case of joint-slice optimization with a single [PITH_FULL_IMAGE:figures/full_fig_p068_2_19.png]
Figure 2.20
Figure 2.20. Figure 2.20: Median values obtained when focusing on the case of joint-slice optimization with a [PITH_FULL_IMAGE:figures/full_fig_p069_2_20.png]
Figure 2.21
Figure 2.21. Figure 2.21: UE Satisfaction expressed in the form of PRB ratio for the Reward Designs of Table 2.11. [PITH_FULL_IMAGE:figures/full_fig_p071_2_21.png]
Figure 2.22
Figure 2.22. Figure 2.22: Impact of the Reward Designs from Table 2.11 on DL eMBB Throughput, DL mMTC [PITH_FULL_IMAGE:figures/full_fig_p072_2_22.png]
Figure 2.23
Figure 2.23. Figure 2.23: Impact of the Reward Designs from Table 2.11 on the median values of DL eMBB [PITH_FULL_IMAGE:figures/full_fig_p072_2_23.png]
Figure 2.24
Figure 2.24. Figure 2.24: Selection of PRB Actions for the three slices from the xApp Catalog of Table 2.12. [PITH_FULL_IMAGE:figures/full_fig_p074_2_24.png]
Figure 2.25
Figure 2.25. Figure 2.25: Resource Utilization and UE Satisfaction for a network deployment described by three [PITH_FULL_IMAGE:figures/full_fig_p075_2_25.png]
Figure 2.26
Figure 2.26. Figure 2.26: Impact of the network deployment described by three different Use Cases on DL eMBB [PITH_FULL_IMAGE:figures/full_fig_p076_2_26.png]
Figure 2.27
Figure 2.27. Figure 2.27: Impact of the network deployment described by three Use Cases on the median values [PITH_FULL_IMAGE:figures/full_fig_p076_2_27.png]
Figure 2.28
Figure 2.28. Figure 2.28: PACIFISTA architecture and workflow [2]. [PITH_FULL_IMAGE:figures/full_fig_p079_2_28.png]
Figure 2.29
Figure 2.29. Figure 2.29: ECDFs of throughput of the eMBB slice for xApps [PITH_FULL_IMAGE:figures/full_fig_p083_2_29.png]
Figure 3.1
Figure 3.1. Figure 3.1: Results averaged over 100 channel realizations for the link describing the green vehicle’s transmission to the red one in Willow Grove Naval Air Base in Horsham Township, Pennsylvania, USA. The implementation of GBSMs is fairly simple and computationally inexpensive,…
Figure 3.2
Figure 3.2. Figure 3.2: UL cellular metrics between BS-node 1 and UE-node 2. 3.1.3 Investigating Energy Efficiency with frequency-dependent GBSMs and RISs in C-V2X Evidently, QuaDRiGa is effective in providing accurate geometry-based stochastic channel modeling. Additionally, considering th…
Figure 3.3
Figure 3.3. Figure 3.3: RIS-assisted and UAV-enabled communications system. [PITH_FULL_IMAGE:figures/full_fig_p095_3_3.png]
Figure 3.4
Figure 3.4. Figure 3.4: A hierarchical Stackelberg game-theoretic approach to energy-efficient power control. [PITH_FULL_IMAGE:figures/full_fig_p103_3_4.png]
Figure 3.5
Figure 3.5. Figure 3.5: Evaluation of the hierarchical game-theoretic power control framework per UE, under [PITH_FULL_IMAGE:figures/full_fig_p104_3_5.png]
Figure 3.6
Figure 3.6. Figure 3.6: Evaluation of the hierarchical game-theoretic power control framework on the entire [PITH_FULL_IMAGE:figures/full_fig_p105_3_6.png]
Figure 3.7
Figure 3.7. Figure 3.7: Path gains for two different spectrum bands. [PITH_FULL_IMAGE:figures/full_fig_p106_3_7.png]
Figure 3
Figure 3. Figure 3: b depicts the path gains obtained for the mmWave band with the carrier frequency [PITH_FULL_IMAGE:figures/full_fig_p107_3.png]
Figure 3.8
Figure 3.8. Figure 3.8: Evaluation of the hierarchical game-theoretic power control framework per UE, under [PITH_FULL_IMAGE:figures/full_fig_p108_3_8.png]
Figure 3.9
Figure 3.9. Figure 3.9: Evaluation of the hierarchical game-theoretic power control framework under two [PITH_FULL_IMAGE:figures/full_fig_p109_3_9.png]
Figure 3.7
Figure 3.7. Figure 3.7: As a next step, we proceed by installing the generated channels in Colosseum. In summary, the reference input scenario to be installed includes two key components: the path gain, represented as a complex coefficient value in the form of FIR, and the Time-of-Arrival (…
Figure 3.10
Figure 3.10. Figure 3.10: Evaluation of a slice-based multi-UE C-V2X communication system on Colosseum [PITH_FULL_IMAGE:figures/full_fig_p111_3_10.png]
Figure 3.11
Figure 3.11. Figure 3.11: Reference O-RAN testing architecture for RIS-enabled deployments, focusing on the [PITH_FULL_IMAGE:figures/full_fig_p119_3_11.png]
Figure 3.12
Figure 3.12. Figure 3.12: Path Gains under varying numbers of RIS elements. [PITH_FULL_IMAGE:figures/full_fig_p122_3_12.png]
Figure 3.13
Figure 3.13. Figure 3.13: Performance evaluation results for the single-UE Case Study. [PITH_FULL_IMAGE:figures/full_fig_p124_3_13.png]
Figure 3.14
Figure 3.14. Figure 3.14: Performance evaluation results for the multi-UE Case Study. [PITH_FULL_IMAGE:figures/full_fig_p125_3_14.png]
Figure 3.15
Figure 3.15. Figure 3.15: Performance evaluation results for the multi-UE O-RAN Case Study. [PITH_FULL_IMAGE:figures/full_fig_p126_3_15.png]
Figure 3.16
Figure 3.16. Figure 3.16: Performance evaluation results for the multi-UE Case Study with and without the [PITH_FULL_IMAGE:figures/full_fig_p126_3_16.png]
Figure 3
Figure 3. Figure 3: a shows that the median eMBB throughput with a RIS in the topology is [PITH_FULL_IMAGE:figures/full_fig_p127_3.png]
Figure 4.1
Figure 4.1. Figure 4.1: System-level simulations using SIONNA over ray-traced channels for link adaptation [PITH_FULL_IMAGE:figures/full_fig_p135_4_1.png]
Figure 4.2
Figure 4.2. Figure 4.2: End-to-end execution of MCS selection in SIONNA. The channel matrix is generated [PITH_FULL_IMAGE:figures/full_fig_p136_4_2.png]
Figure 4.3
Figure 4.3. Figure 4.3: Mean Spectral Efficiency and BLER for all RL methods and baselines under Setup A [PITH_FULL_IMAGE:figures/full_fig_p138_4_3.png]
Figure 4.4
Figure 4.4. Figure 4.4: Distribution of Selected MCS Indices across Methods and UEs under Setup A from [PITH_FULL_IMAGE:figures/full_fig_p139_4_4.png]
Figure 4
Figure 4. Figure 4: shows the relative frequency distribution of selected MCS indices across different [PITH_FULL_IMAGE:figures/full_fig_p140_4.png]
Figure 4.5
Figure 4.5. Figure 4.5: CDFs of Spectral Efficiency, BLER, and MCS under Setup A from Table 4.1, leveraging [PITH_FULL_IMAGE:figures/full_fig_p141_4_5.png]
Figure 4.6
Figure 4.6. Figure 4.6: Spectral Efficiency and BLER under Setup A from Table 4.1, using the DT predictor [PITH_FULL_IMAGE:figures/full_fig_p142_4_6.png]
Figure 4.7
Figure 4.7. Figure 4.7: Mean Spectral Efficiency and BLER for all RL methods and baselines under Setup A [PITH_FULL_IMAGE:figures/full_fig_p142_4_7.png]
Figure 4.8
Figure 4.8. Figure 4.8: Training reward over episodes [PITH_FULL_IMAGE:figures/full_fig_p143_4_8.png]
Figure 4.9
Figure 4.9. Figure 4.9: Offline FQI results on OTA data for UL (top) and DL (bottom). Left: Q-value convergence [PITH_FULL_IMAGE:figures/full_fig_p144_4_9.png]
Figure 5.1
Figure 5.1. Figure 5.1: Interference from terrestrial cellular DL transmissions to satellite UL in terrestrial-LEO [PITH_FULL_IMAGE:figures/full_fig_p150_5_1.png]
Figure 5.2
Figure 5.2. Figure 5.2: End-to-end resource management framework overview. [PITH_FULL_IMAGE:figures/full_fig_p155_5_2.png]
Figure 5.3
Figure 5.3. Figure 5.3: Simulation topology for terrestrial-satellite spectrum sharing. [PITH_FULL_IMAGE:figures/full_fig_p155_5_3.png]
Figure 5.4
Figure 5.4. Figure 5.4: Satellite overview with QuaDRiGa’s visualize lotlan in various geographic regions. The trajectories shown correspond to different parameterizations of the orbital elements defined in Listing 5.1. 1 h_qd_sat = qd_satellite(constellation, Ain, Bin, Cin, Din, Ein, Fin);…
Figure 5.5
Figure 5.5. Figure 5.5: Probability Density Function (PDF) of the satellite elevation angles for the constellation [PITH_FULL_IMAGE:figures/full_fig_p158_5_5.png]
Figure 5.6
Figure 5.6. Figure 5.6: Beamforming gain of the considered gNB with a [PITH_FULL_IMAGE:figures/full_fig_p161_5_6.png]
Figure 5
Figure 5. Figure 5: illustrates the resulting transmit beamforming gains [PITH_FULL_IMAGE:figures/full_fig_p162_5.png]
Figure 5.7
Figure 5.7. Figure 5.7: CDF of the INR as defined in (5.6): Power Control & Standard Beamforming vs. Interference Nulling. Interf. Nulling (λ = 0) Interf. Nulling (λ = 0.1) Interf. Nulling (λ = 1) Interf. Nulling (λ = 10) No Null., Power Control 0 10 20 RSS Degradation [dB] (a) 0 1 2 3 0 10…
Figure 5.8
Figure 5.8. Figure 5.8: Worst-case evaluation and fairness analysis for the terrestrial network topology described [PITH_FULL_IMAGE:figures/full_fig_p163_5_8.png]
Figure 5
Figure 5. Figure 5: (c). We observe that [PITH_FULL_IMAGE:figures/full_fig_p164_5.png]
Figure 5.9
Figure 5.9. Figure 5.9: Utility function analysis for the terrestrial network topology described in Table 5.1. [PITH_FULL_IMAGE:figures/full_fig_p165_5_9.png]
Figure 5.10
Figure 5.10. Figure 5.10: Achieved utility and transmit power selection for the gNB DL under varying terrestrial [PITH_FULL_IMAGE:figures/full_fig_p166_5_10.png]
Figure 5.11
Figure 5.11. Figure 5.11: CDF of the INR as defined in (5.6) for all optimization strategies. [PITH_FULL_IMAGE:figures/full_fig_p167_5_11.png]
Figure 5.12
Figure 5.12. Figure 5.12: Worst-case RSS degradation as defined in (5.8) for all optimization strategies. [PITH_FULL_IMAGE:figures/full_fig_p168_5_12.png]
Figure 5.13
Figure 5.13. Figure 5.13: Performance evaluation of joint interference nulling and power control: (a) INR, (b) [PITH_FULL_IMAGE:figures/full_fig_p169_5_13.png]
Figure 5.14
Figure 5.14. Figure 5.14: CDF of RSS degradation for λ = 1 with varying number of satellites. −60 −40 −20 0 20 40 60 −80 −60 −40 −20 0 20 40 60 80 1 2 3 4 5 6 7 8 9 10 11 12 14 13 15 16 17 19 18 20 21 22 23 24 25 26 27 28 29 30 Azimuth angle (ϕ) [◦ ] Elevation angle ( θ) [◦ ] Satellites UEs …
Figure 5.15
Figure 5.15. Figure 5.15: Angular positions for the complete topology for different antenna sizes with [PITH_FULL_IMAGE:figures/full_fig_p170_5_15.png]
Figure 5.16
Figure 5.16. Figure 5.16: Channel analysis: (a) Average signal leakage towards all satellites, and (b) UE RSS [PITH_FULL_IMAGE:figures/full_fig_p171_5_16.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

19 extracted references

  1. [248]

    Gymnasium: A Standard Interface for Reinforcement Learning Environ- ments,

    Farama Foundation, “Gymnasium: A Standard Interface for Reinforcement Learning Environ- ments,” GitHub Repository, 2023, [Available Online]: https://github.com/Farama-Foundation/ Gymnasium

    2023

  2. [249]

    Induction of decision trees,

    J. R. Quinlan, “Induction of decision trees,”Machine Learning, vol. 1, no. 1, pp. 81–106, 1986

    1986

  3. [250]

    A New Approach to Linear Filtering and Prediction Problems,

    R. E. Kalman, “A New Approach to Linear Filtering and Prediction Problems,”Transactions of the ASME–Journal of Basic Engineering, vol. 82, pp. 35–45, 1960

    1960

  4. [251]

    Random Forests,

    L. Breiman, “Random Forests,”Machine Learning, vol. 45, 2001

    2001

  5. [252]

    Tree-based batch mode reinforcement learning,

    D. Ernst, P. Geurts, and L. Wehenkel, “Tree-based batch mode reinforcement learning,” Journal of Machine Learning Research, vol. 6, 2005

    2005

  6. [253]

    Spectrum sharing and interference management for 6G LEO satellite-terrestrial network integration,

    N. Heydarishahreza, T. Han, and N. Ansari, “Spectrum sharing and interference management for 6G LEO satellite-terrestrial network integration,”Communications Surveys & Tutorials, 2024

    2024

  7. [254]

    ASCENT: A Context-Aware Spectrum Coexistence Design and Implementation Toolset for Policymakers in Satellite Bands,

    T.-S. R. Niloy, S. Kumar, A. Hore, Z. Hassan, C. Dietrich, E. W. Burger, J. H. Reed, and V . K. Shah, “ASCENT: A Context-Aware Spectrum Coexistence Design and Implementation Toolset for Policymakers in Satellite Bands,” inInternational Symposium on Dynamic Spectrum Access Networks (DySPAN). IEEE, 2024, pp. 240–248

    2024

  8. [255]

    Interference Analysis of Coexisting 5G Networks and NGSO FSS Receivers in the 12-GHz Band,

    T.-S. R. Niloy, Z. Hassan, N. Stephenson, and V . K. Shah, “Interference Analysis of Coexisting 5G Networks and NGSO FSS Receivers in the 12-GHz Band,”IEEE Wireless Communications Letters, vol. 12, no. 9, pp. 1528–1532, 2023

    2023

Show all 19 references
  1. [256]

    Space-Ground Multicast Group Control for Multiuser LEO Satellite Networks,

    D. Wu, C. Qin, Y . Cui, P. He, and R. Wang, “Space-Ground Multicast Group Control for Multiuser LEO Satellite Networks,”IEEE Transactions on Wireless Communications, 2023

    2023

  2. [257]

    Resource Scheduling for High- Capacity Multicast Service in Ultra-Dense LEO Satellite Networks,

    T. Ma, B. Qian, X. Qin, X. Zhang, L. X. Cai, and H. Zhou, “Resource Scheduling for High- Capacity Multicast Service in Ultra-Dense LEO Satellite Networks,”IEEE Transactions on Vehicular Technology, 2023

    2023

  3. [258]

    Satellite-Terrestrial Coexistence in FR3 Band via Hybrid True- Time-Delay Array-based Nulling,

    A. Wadaskar and D. Cabric, “Satellite-Terrestrial Coexistence in FR3 Band via Hybrid True- Time-Delay Array-based Nulling,” inIEEE International Symposium on Dynamic Spectrum Access Networks, 2025. 184

    2025

  4. [259]

    Performance and interference criteria for satellite passive remote sensing,

    ITU-R RS.2017-0, “Performance and interference criteria for satellite passive remote sensing,” International Telecommunication Union, Radiocommunication Sector, Recommendation, 2012

    2017

  5. [260]

    Study on channel model for frequencies from 0.5 to 100 GHz,

    G. T. 38.901, “Study on channel model for frequencies from 0.5 to 100 GHz,” 3GPP, Technical Report, 2023

    2023

  6. [261]

    A 5G-NR satellite extension for the QuaDRiGa channel model,

    S. Jaeckel, L. Raschkowski, and L. Thieley, “A 5G-NR satellite extension for the QuaDRiGa channel model,” inIEEE Joint European Conference on Networks and Communications & 6G Summit, 2022

    2022

  7. [262]

    An analytical model to predict the probability density function of elevation angles for LEO satellite systems,

    S.-Y . Li and C. Liu, “An analytical model to predict the probability density function of elevation angles for LEO satellite systems,”IEEE Communications Letters, vol. 6, no. 4, 2002

    2002

  8. [263]

    A quantitative measure of fairness and discrimination,

    R. K. Jain, D.-M. W. Chiu, W. R. Haweet al., “A quantitative measure of fairness and discrimination,”Eastern Research Laboratory, Digital Equipment Corporation, Hudson, MA, vol. 21, no. 1, 1984

    1984

  9. [264]

    MATLAB (R2025a),

    MathWorks, “MATLAB (R2025a),” [Available Online]: https://www.mathworks.com, 2025

    2025

  10. [265]

    C. A. Balanis,Antenna theory: analysis and design. John Wiley & Sons, 2016

    2016

  11. [266]

    H. L. Van Trees,Optimum array processing: Part IV of detection, estimation, and modulation theory. John Wiley & Sons, 2002. 185 Acknowledgments The completion of my Ph.D. brings back so many memories of hope, perseverance, joy, effort, and determination to succeed. The past fi...

    2002

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.