cs.ITOct 15, 2026

Uncertainty-Aware Federated Learning for Joint Communication and Computation in XR Systems

Authors: Robert-Jeron Reifert, Hayssam Dahrouj, Alaa Alameer Ahmad, Haris Gacanin, Aydin Sezgin

Organizations: Digital Communication Systems, Ruhr University Bochum, Bochum, Germany. · the Department of Electrical Engineering, University of Sharjah, Sharjah, United Arab Emirates. · Cariad SE, Wolfsburg, Germany. · the Chair for Distributed Signal Processing, RWTH Aachen University, Aachen, Germany.

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

The convergence of communication and computation, empowered by artificial intelligence (AI) enhanced capabilities, emerges as a key enabler for realizing the promising functionalities of the sixth-generation of communication systems (6G). To this end, this paper investigates an AI-native, robust network, where multiple base stations (BSs) simultaneously serve several extended reality (XR) devices. The paper then presents an innovative distributed deep learning-based approach to simultaneously manage the transmit and computing powers, and computation allocation, amidst uncertainties in both channel and computing states information. Under such an interference network scenario, the paper proposes a robust approach for minimizing the worst-case delay across the served XR devices, subject to computation and power constraints. The paper leverages a deep neural network (DNN) that can be trained either centrally or in a federated manner to map estimated channels and computation requirements to optimized resource allocations per BS. During training, uncertainty samples, reflecting both channel and computing estimation errors, are injected after the DNN output to account for such joint uncertainties. Our results validate the enhanced robust delay performance and scalability of joint uncertainty injection as compared to uncertainty-agnostic methods and state-of-the-art benchmarks, with federated learning-based solutions achieving near-centralized performance while enhancing efficiency, privacy, and adaptability.

Figures & tables

Fig. 1 : Network setup with one central server and global DNN model, 33 multi antenna BSs each serving a number of XR devices accommodating a local DNN model. Devices’ communication, computation, and inter-BS interference links are shown.
Fig. 2 : Proposed DNN architecture with estimated input, resource allocation output, and uncertainty injection. Modifications for the distributed DNN are highlighted in green.
Fig. 3 : Channel knowledge representation for 2 BSs and 5 devices illustrating known and unknown channels.
GeneralCommunicationComputation
Pmax=34P^{\text{max}}=34 dBmW=10W=10 MHzFmax=4.6⋅109F^{\text{max}}=4.6\cdot 10^{9} cycles/s
N=1000N=1000 samplesDkout=7.5⋅104D_{k}^{\text{out}}=7.5\hskip-1.42271pt\cdot\hskip-1.42271pt10^{4} bitsDkin=5⋅104D_{k}^{\text{in}}=5\cdot 10^{4} bits
γ=0.05\gamma=0.05α=0.2\alpha=0.2 RZF factorτ=10−28\tau=10^{-28} , μ=3\mu=3
 σ2=−175\sigma^{2}=-175 dBm/Hzωk∼Γ(2,200)\omega_{k}\sim\Gamma(2,200) , ∀k∈K\forall k\in\mathcal{K}
TABLE I : Simulation parameters (unless otherwise mentioned).
Fig. 4 : Robust delay t^maxγ\hat{t}_{\text{max}}^{\hskip 0.56917pt\gamma} over channel uncertainty, i.e., σh2\sigma_{h}^{2} .
Fig. 5 : Robust delay t^maxγ\hat{t}_{\text{max}}^{\hskip 0.56917pt\gamma} over computing uncertainty, i.e., σω2\sigma_{\omega}^{2} .
Fig. 6 : Robust delay t^maxγ\hat{t}_{\text{max}}^{\hskip 0.56917pt\gamma} over computing uncertainty, i.e., σω2\sigma_{\omega}^{2} .
Fig. 7 : Robust worst-case delay t^maxγ\hat{t}_{\text{max}}^{\hskip 0.56917pt\gamma} vs. number of devices KK .
Fig. 8 : Robust delay t^maxγ\hat{t}_{\text{max}}^{\hskip 0.56917pt\gamma} over network density.
Network densitysparsemoderatedense
Unit in mstinfdnn\bm{t^{\text{dnn}}_{\text{inf}}}t^maxγ\bm{\hat{t}_{\text{max}}^{\hskip 0.56917pt\gamma}}tinfdnn\bm{t^{\text{dnn}}_{\text{inf}}}t^maxγ\bm{\hat{t}_{\text{max}}^{\hskip 0.56917pt\gamma}}tinfdnn\bm{t^{\text{dnn}}_{\text{inf}}}t^maxγ\bm{\hat{t}_{\text{max}}^{\hskip 0.56917pt\gamma}}
Non-robust FL4.49843.7174.30191.9184.525144.651
Robust FL4.40831.3924.19965.3984.288112.939
Non-robust FP8.5 ⋅103\cdot 10^{3}59.84712.2 ⋅103\cdot 10^{3}129.59220.2 ⋅103\cdot 10^{3}212.580
SAA/WMMSE9.6 ⋅103\cdot 10^{3}42.52038.6 ⋅103\cdot 10^{3}96.609104.6 ⋅103\cdot 10^{3}176.476
Uniform FP6.4 ⋅103\cdot 10^{3}42.5829.9 ⋅103\cdot 10^{3}100.18716.9 ⋅103\cdot 10^{3}178.833
TABLE II : Mean inference/optimization time tinfdnn\bm{t^{\text{dnn}}_{\text{inf}}} for ML-based schemes and optimization baselines.
Fig. 9 : Robust worst-case delay t^maxγ\hat{t}_{\text{max}}^{\hskip 0.56917pt\gamma} vs. data heterogeneity for different FL implementations.
Fig. 10 : Robust worst-case delay t^maxγ\hat{t}_{\text{max}}^{\hskip 0.56917pt\gamma} vs. parameter γ\gamma , showing impact of hard latency deadline guarantees.
Fig. 11 : Robust worst-case delay t^maxγ\hat{t}_{\text{max}}^{\hskip 0.56917pt\gamma} vs. three uncertainty levels ({ low , medium , high } corresponding to (σh2,σω2)∈{(0.004,800),(0.008,1600),(0.016,3200)}(\sigma_{h}^{2},\sigma_{\omega}^{2})\in\{(0.004,800),(0.008,1600),(0.016,3200)\} , respectively).
Fig. 12 : Convergence behavior of centralized learning and FL.
Fig. 13 : Robust worst-case delay t^maxγ\hat{t}_{\text{max}}^{\hskip 0.56917pt\gamma} vs. training epochs showing the impact of uncertainty samples NN .
Training Epochs E\mathbf{E}200\mathbf{200}300\mathbf{300}400\mathbf{400}500\mathbf{500}
N=500N=50044296657888511110
N=1000N=10008593129061722221533
TABLE III : Training time in seconds for Robust FL .
Fig. 14 : Exchanged data during training in MegaByte (MB) versus the number of BSs for different device numbers.
Fig. 15 : Exchanged data for inference in kiloByte (kB) versus the number of BSs for different device numbers.
Fig. 16 : Robust worst-case delay t^maxγ\hat{t}_{\text{max}}^{\hskip 0.56917pt\gamma} vs. neurons per hidden layer nhn_{h} (left) and number of hidden layers ηh\eta_{h} (right).
More in this topic

Wireless Resource Allocation

Wireless resource allocation assigns spectrum, power, beams, and computing capacity to meet throughput, latency, fairness, and energy goals.

See all 172 papers

Explore similar work

CardsList
Mar 10, 2026

Agentic AI as a Network Control-Plane Intelligence Layer for Federated Learning over 6G

The shift toward user-customized on-device learning places new demands on wireless systems: models must be trained on diverse, distributed data while meeting strict latency, bandwidth, and reliability constraints. To address this, we propose an Agentic AI as the control layer for managing federated learning (FL) over 6G networks, which translates high-level task goals into actions that are aware of network conditions. Rather than simply viewing FL as a learning challenge, our system sees it as a combined task of learning and network management. A set of specialized agents focused on retrieval, planning, coding, and evaluation utilizes monitoring tools and optimization methods to handle client selection, incentive structuring, scheduling, resource allocation, adaptive local training, and code generation. The use of closed-loop evaluation and memory allows the system to consistently refine its decisions, taking into account varying signal-to-noise ratios, bandwidth conditions, and device capabilities. Finally, our case study has demonstrated the effectiveness of the Agentic AI system's use of tools for achieving high performance.

Loc X. Nguyen, Ji Su Yoon, Huy Q. Le +6

Department of Computer Science and Engineering, Kyung Hee University, Yongin-si, Gyeonggi-do 17104, Republic of Korea. · Department of Artificial Intelligence, Kyung Hee University, Yongin-si, Gyeonggi-do 17104, Republic of Korea. · School of Computer Science, The University of Sydney, Australia +1

Federated Learning6G Networks
AskPDF
Oct 5, 2026

Combining Improvements in Uplink AI-RAN

One of the major transformative factors in 6G will be the integration of Artificial Intelligence (AI) to become a native part of Radio Access Network (RAN). While most physical-layer AI features have so far been evaluated in isolation using link-level simulations, their combined behavior in a realistic multi-cell, multi-UE deployment has remained largely unexplored. In this paper, we present system-level performance results when multiple uplink AI features are enabled together, achieved by integrating accurate link-level and system-level simulators. To infer state-of-the-art deep-learning-aided Multiple Input Multiple Output (MIMO) receivers under the dynamic allocations produced by a realistic uplink scheduler, we propose a mirrored data augmentation method that decouples receiver performance from scheduled allocation size. In addition to these Physical Layer (PHY) receiver features, we combine several recent advances in deep reinforcement learning to train uplink power control and link adaptation that outperform a heuristic baseline and further boost the gains obtainable from the AI receiver alone. The system-level results show that the combined AI features improve the mean uplink user throughput by roughly 27% compared to a non-AI baseline, confirming that the individual PHY and Medium Access Control (MAC) AI features provide complementary gains when deployed jointly.

Petteri Kela, Dani Korpi, Mikko Honkala

Nokia Technologies Espoo, Finland · Nokia Bell Labs Espoo, Finland

Wireless CommunicationsWireless Resource Allocation
AskPDF
Feb 24, 2026

Federated Multi Agent Deep Learning and Neural Networks for Advanced Distributed Sensing in Wireless Networks

Multi-agent deep learning (MADL), including multi-agent deep reinforcement learning (MADRL), distributed/federated training, and graph-structured neural networks, is becoming a unifying framework for decision-making and inference in wireless systems where sensing, communication, and computing are tightly coupled. Recent 5G-Advanced and 6G visions strengthen this coupling through integrated sensing and communication, edge intelligence, open programmable RAN, and non-terrestrial/UAV networking, which create decentralized, partially observed, time-varying, and resource-constrained control problems. This survey synthesizes the state of the art, with emphasis on 2021-2025 research, on MADL for distributed sensing and wireless communications. We present a task-driven taxonomy across (i) learning formulations (Markov games, Dec-POMDPs, CTDE), (ii) neural architectures (GNN-based radio resource management, attention-based policies, hierarchical learning, and over-the-air aggregation), (iii) advanced techniques (federated reinforcement learning, communication-efficient federated deep RL, and serverless edge learning orchestration), and (iv) application domains (MEC offloading with slicing, UAV-enabled heterogeneous networks with power-domain NOMA, intrusion detection in sensor networks, and ISAC-driven perceptive mobile networks). We also provide comparative tables of algorithms, training topologies, and system-level trade-offs in latency, spectral efficiency, energy, privacy, and robustness. Finally, we identify open issues including scalability, non-stationarity, security against poisoning and backdoors, communication overhead, and real-time safety, and outline research directions toward 6G-native sense-communicate-compute-learn systems.

Nadine Muller, Stefano DeRosa, Su Zhang +1

Faculty of Information Technology and Electrical Engineering, University of Oulu

Multi-Agent Reinforcement LearningMulti-Agent SystemsFederated RLIntegrated Sensing and Communication
AskPDF