PaperlayerView exact revision on arXiv

Source-linked AI summary

Perception-Aware Joint Power and Sub-Band Allocation for 6G In-Body Subnetworks

Samira Abdelrahman, Hossam Farag

arXiv:2609.07519v1cs.NI

TL;DR

Dense IBS coexistence can make meeting application-level latency targets power-intensive, while those targets do not account for users’ delay perception. This paper learns user perceptual thresholds and incorporates them into joint power and sub-band allocation. Simulations report power reductions of approximately 26% under stringent latency requirements and savings approaching 60% in dense IBS deployments while maintaining required QoE.

  • Problem

    IBS interference-management approaches overlook user perception, even though latency reductions below a user’s perceptual threshold may not improve QoE and can make resource use inefficient.

    Source: §I · pp. 1–2

  • Method

    The paper estimates user-specific JNDs with GMM clustering and supervised learning, then incorporates the learned model into perception-aware joint power and sub-band allocation.

    Source: §I · p. 2, A. Modeling and Estimation of the JND · p. 3

  • Results

    At Dmax 10 ms, transmit power falls from 9.75 dBm to 8.44 dBm (approximately 26% less); at N = 30, power savings are approximately 60%.

    Source: §IV · p. 6

  • Takeaways & Limitations

    Simulations indicate that perception-aware allocation can reduce transmit power while maintaining required QoE, with savings approaching 60% in dense IBS deployments.

    Source: §I · p. 2, §IV · p. 6

Abstract

from arXiv · show

In-body subnetworks (IBSs) are expected to become a key enabler of immersive eXtended Reality (XR) services in sixth-generation (6G) networks by providing ultra-short-range, low-latency wireless connectivity around the human body. However, the dense coexistence of multiple IBSs leads to severe co-channel interference, requiring increased transmit power to satisfy the stringent latency requirements of XR applications. Existing interference management approaches allocate radio resources solely according to application-level Quality-of-Service (QoS) requirements, overlooking the perceptual limitations of human users. This paper proposes perception-aware joint power control and sub-band allocation framework that integrates users' delay perception into radio resource allocation for XR-oriented IBSs. A learning-based perception model is first developed by combining Gaussian mixture modeling (GMM) with supervised learning to develop a statistical model of the delay perception threshold. The learned perception model is then incorporated into a stochastic radio resource allocation problem, which is reformulated using a Lyapunov drift-plus-penalty and solved through a low-complexity per-slot resource allocation procedure. System-level simulations under realistic intra- and inter-IBS propagation conditions demonstrate that the proposed approach substantially improves radio resource efficiency, achieving up to 26% transmit power reduction under stringent latency requirements and approximately 60% power savings in dense IBS deployments, while maintaining the required Quality of Experience (QoE).

I. INTRODUCTION

IBSs support short-range XR connectivity, but growing coexistence makes interference costly under application-level latency targets. The paper proposes allocating power and sub-bands with users’ delay perception in view.

I. INTRODUCTION

  • Dense IBS coexistence makes interference limit reliable high-rate XR video delivery and can drive transmit power increases to meet application-level QoS constraints.IBSs are short-range wireless cells near the body, and XR is a representative use case.Source: §I · p. 1
  • Prior work addresses interference in industrial subnetworks through centralized, deep-learning, multi-agent, adaptive, and decentralized resource-management approaches.The cited approaches target objectives such as interference reduction, data-rate satisfaction, reliability, spectrum use, or coordination of application and control goals.Source: §I · p. 1
  • Existing IBS interference-management approaches largely neglect user perception, although latency improvements below a perceptual threshold may not improve experienced QoE.This can make additional power used to improve imperceptible QoS gains inefficient.Source: §I · pp. 1–2
  • The paper proposes a user-specific, time-varying just-noticeable delay model and uses learned delay perception in power and sub-band allocation.The framework combines GMM clustering with supervised learning and uses Lyapunov-based optimization; simulations report up to approximately 26% less transmit power under stringent latency requirements and savings approaching 60% in dense IBS deployments while maintaining required QoE.Source: §I · p. 2

II. SYSTEM MODEL AND PROBLEM FORMULATION

The system model describes downlink XR video delivery across multiple IBSs sharing sub-bands, with delay shaped by queuing, transmission, and interference. The optimization minimizes long-term transmit power subject to perception-based QoE constraints and feasible power and sub-band assignments.

II. SYSTEM MODEL AND PROBLEM FORMULATION

  • The modeled indoor XR system has multiple IBSs, each with an AP and XR devices, and the APs transmit downlink video frames to associated XR displays.The IBSs share bandwidth divided into sub-bands and operate in time slots.Source: §II · p. 2
  • User delay combines queuing and transmission delay, while achievable rate depends on sub-band SINR and interference from co-scheduled IBS users.The model assumes M/M/1 AP buffers and sub-band bandwidth below the channel coherence bandwidth.Source: §II · pp. 2–3
  • The allocation problem minimizes average AP transmit power while satisfying probabilistic QoE requirements tied to each user’s JND and enforcing feasible power and binary sub-band assignments.The formulation accounts for users’ perceptual delay limits rather than relying only on fixed application-centric latency constraints.Source: §II · p. 3

A. Modeling and Estimation of the JND

The JND estimator combines GMM clustering of user features and delay thresholds with supervised classification of perceptual modes. The identified mode supports a probabilistic model of each user’s delay perception.

A. Modeling and Estimation of the JND

  • The estimator maps user-specific feature vectors, such as age, location, or XR experience, to JNDs using a reliability-aware supervised learning framework.The feature-to-JND data may come from experiments or user studies.Source: A. Modeling and Estimation of the JND · p. 3
  • A GMM fitted with EM estimates the joint distribution of user features and JNDs, identifying perceptual-mode clusters used as labels for supervised learning.The model treats human delay perception as multimodal and assigns each data point to a mode using posterior probability.Source: A. Modeling and Estimation of the JND · pp. 3–4
  • A supervised classifier predicts each user’s perceptual mode from observed features, enabling a probabilistic JND model and an effective delay threshold.The effective delay is described as the minimum delay remaining imperceptible with probability 1 − ϵ′.Source: A. Modeling and Estimation of the JND · p. 4

B. Optimal Resource allocation via Lyapunov optimization

The stochastic delay-reliability constraint is reformulated as a time-average condition that can be handled with Lyapunov drift-plus-penalty optimization. The resulting per-slot procedure accounts for changing channels and user perception while targeting queue stability and lower transmit power.

B. Optimal Resource allocation via Lyapunov optimization

IV. PERFORMANCE EVALUATION

Simulations compare perception-aware allocation with a fixed-delay baseline across latency and network-load settings. The proposed approach uses less transmit power, with the largest reported savings in dense deployments.

IV. PERFORMANCE EVALUATION

  • The system-level setup uses 100 MHz across four sub-bands and distinct intra- and inter-IBS propagation models, treating IBS cylinders as potential inter-body link blockages.IBS locations follow a Thomas Cluster Process; inter-IBS links use 3GPP D2D propagation models.Source: §IV · p. 5
  • Delay-perception estimates use ratings from 30 participants evaluating five videos, while user features are synthetically generated because public paired feature and perception data are unavailable.The synthesized features have a GMM marginal distribution and are statistically correlated with individual delay perception.Source: §IV · pp. 5–6
  • Both schemes use less average transmit power as Dmax_i increases; at 60 ms, savings become negligible as Dmax_i approaches the JND threshold.The perception-aware advantage is most pronounced under stringent latency constraints.Source: §IV · p. 6
  • At N=30, total average transmit power is roughly 4 dBm lower than baseline (about 60% savings), versus approximately 1 dBm at N=2.The comparison is shown as network size increases with Dmax_i = 20 ms.Source: Figure 3 · p. 6, §IV · p. 6
  • For two users, the 156.56 ms JND user receives less power than the 30.2 ms JND user; baseline users’ power differences mainly reflect physical channel gains.The two users follow the proposed method, while the other eight IBSs use predefined delay requirements.Source: Figure 4 · p. 6, §IV · p. 7
  • At Dmax_i = 10 ms, the proposed scheme reduces transmit power from 9.75 dBm to 8.44 dBm, about 26% less than baseline.The advantage is attributed to using learned JNDs to avoid resources spent on imperceptible QoS gains.Source: §IV · p. 6, Figure 2 · p. 6

V. CONCLUSION

The paper integrates users’ delay perception into XR-oriented IBS resource allocation and reports lower transmit power while maintaining acceptable QoE. Experimental validation with real users and measured delay-perception data remains a future extension.

V. CONCLUSION

  • The framework combines delay-perception estimation with joint power and sub-band allocation for XR-oriented 6G IBSs.The learned perceptual threshold is integrated into a Lyapunov drift-plus-penalty allocation framework.Source: §V · p. 7
  • Simulations show reduced average transmit power while maintaining acceptable QoE in challenging, dense deployment scenarios.Source: §V · p. 7
  • Practical benefits remain to be assessed through experimental validation with real XR users and measured delay-perception datasets.Source: §V · p. 7
Loading 2609.07519v1…