mcs adaptation

This work addresses the challenge of link adaptation in wireless communications, where the goal is to dynamically select the optimal modulation and coding scheme (MCS) to maximize throughput. Conventional multi-armed bandit (MAB) approaches often exhibit unstable performance because they neglect the inherent monotonicity in MCS success probabilities—higher-order MCSs typically yield lower success rates under identical channel conditions. To overcome this limitation, the paper formulates link adaptation as an MAB problem and introduces a novel Thompson sampling algorithm that employs a joint ordered Beta distribution as its prior, explicitly preserving the monotonic structure among MCS success rates. By integrating both ACK/NACK and channel quality indicator (CQI) feedback, the proposed method achieves consistently high and competitive throughput across diverse channel scenarios, significantly outperforming existing MAB-based approaches.

To address the inefficiency and throughput limitations of ARQ/HARQ retransmission mechanisms under high packet-loss conditions in 5G networks, this paper proposes and implements a forward erasure coding scheme based on Random Linear Network Coding (RLNC) at the IP layer. Leveraging the netfilter framework, the scheme intercepts and encodes packets in real time over the gNB–UE wireless link. Crucially, it replaces conventional retransmissions with RLNC-based encoding at the block level, thereby eliminating dependency on feedback channels. Experimental results demonstrate that, under moderate code rates, the scheme significantly reduces retransmission counts and associated resource overhead. Throughput improvements reach 30%–60% across medium-to-high packet-loss regimes, while jitter remains within acceptable bounds. These findings validate RLNC as a feasible and practical paradigm for achieving efficient and reliable data transmission in 5G systems.

This paper addresses performance degradation in short-block-length transmission systems caused by unknown channel state information (CSI) and low-density pilot signals. We propose a joint detection and channel estimation framework for bit-interleaved coded modulation (BICM). Our key contributions are: (1) a novel joint BICM metric enabling end-to-end joint decoding and estimation assisted by training signals; (2) the first demonstration in OFDM systems of near-ideal coherent reception performance with only a 4-symbol detection window; and (3) an adaptive Demodulation Reference Signal (DMRS) power allocation scheme that jointly optimizes channel estimation accuracy and coding gain under low-overhead constraints. Evaluated on a full 5G link—featuring Polar/LDPC coding, BPSK/QPSK modulation, and OFDM—the scheme achieves significantly lower bit error rates (BER) than conventional separate-receiver architectures for ultra-short blocks (<64 bits), delivering up to 1.8 dB coding gain and approaching the perfect-CSI performance bound even with sparse DMRS placement.

This paper addresses the path selection problem from edge users to the core network in wireless mesh networks. We propose an interference-aware tree-search path optimization algorithm designed to maximize the end-to-end signal-to-noise-plus-interference ratio (SNIR). Unlike conventional approaches that neglect interference, our method is the first to embed a global wireless interference model directly into the tree-search framework—thereby preserving solution quality while substantially reducing computational complexity. Experimental evaluation across three mesh network scales demonstrates that the selected paths achieve a minimum SNIR 3–18 dB higher than those produced by interference-agnostic methods, and 16–20 dB and 0.5–7 dB higher than those of random and genetic algorithms, respectively; moreover, our algorithm incurs significantly lower runtime overhead. This work establishes a new paradigm for efficient, scalable path assignment in high-interference edge networking environments.

URLLC-oriented IoT resource allocation faces a fundamental trade-off among ultra-low latency, high reliability, and computational complexity. To address this, we propose a dynamic CSI-driven lightweight graph neural network (GNN) framework for resource allocation. Our method introduces a novel dynamic pilot allocation mechanism to adaptively ensure CSI freshness and extends GNN-based modeling to time-varying channel environments by jointly incorporating CSI temporal correlation and greedy heuristic optimization. Evaluated in dense IoT networks, the proposed approach improves spectral efficiency by over 12% compared to conventional greedy algorithms; integrating dynamic pilot allocation yields an additional 3–5% gain. Moreover, it significantly enhances user fairness and system throughput while maintaining low computational overhead, strong scalability, and real-time feasibility—thereby achieving a balanced compromise among performance, timeliness, and practical deployability.

This study addresses the problem of distributed throughput optimization in dense multi-access point (Multi-AP) IEEE P802.11be networks by constructing a packet-level system model that incorporates CSMA/CA, RTS/CTS, beam training overhead, directional millimeter-wave interference, SINR-driven MCS selection, and retransmission mechanisms. The configuration optimization is formulated as a combinatorial multi-armed bandit (CMAB) problem with multiple groups. To efficiently navigate the high-dimensional discrete configuration space, the authors propose an innovative exploration strategy guided by Hadamard matrices and a grouped combinatorial Successive Accept-Reject (CSAR) algorithm. Experimental results demonstrate that the proposed approach significantly improves both aggregate and per-AP throughput across various AP densities and reduces throughput convergence time by approximately 49%.

This work addresses the joint optimization of modulation format, symbol rate, pulse-shaping roll-off factor, and WSS bandwidth in open and disaggregated optical systems under fixed spectrum allocation and limited transceiver resources, aiming to maximize throughput while ensuring a quality-of-transmission (QoT) margin. For the first time, experimental results reveal the critical impact of the roll-off factor on QoT under cascaded WSS filtering, which is then incorporated into a knapsack-based optimization model tailored for Optical Spectrum-as-a-Service (OSaaS) to enable adaptive trade-offs between throughput and QoT margin. Validation on a metro-scale testbed demonstrates that the proposed approach significantly enhances spectral efficiency, achieving an effective balance between throughput and transmission quality.

This study addresses the insufficient reliability of existing semantic communication systems in dynamically uncertain environments, particularly their poor tail performance under harsh channel conditions. For the first time, it introduces a reliability-oriented perspective into semantic communication design and proposes a sample-level reliability enhancement framework that integrates channel-aware adaptation, robust semantic encoding and decoding, and HARQ retransmission mechanisms. To cope with imperfect channel state information, the work further develops a robust adaptive transmission scheme and a joint source-channel-check coding method. The paper systematically categorizes three design paradigms for reliable semantic communication and their limitations, then presents two novel solutions, offering both theoretical foundations and optimization frameworks for building high-reliability semantic communication systems compatible with existing wireless networks.

In wireless link adaptation, it is challenging to jointly optimize spectral efficiency and block error rate (BLER) reliability. Method: This paper proposes SALAD—a novel adaptive algorithm that (i) estimates SINR online via cross-entropy loss minimization using ACK/NACK feedback; (ii) employs knowledge distillation for adaptive learning-rate tuning; (iii) integrates hypothesis testing to accelerate MCS selection under abrupt channel variations; and (iv) establishes a closed-loop feedback mechanism to dynamically adjust the instantaneous BLER target, thereby stabilizing long-term BLER. Contribution/Results: Unlike conventional outer-loop adaptation, SALAD eliminates manual parameter tuning. Evaluated on real 5G deployments, it achieves up to 15% gains in spectral efficiency and throughput while precisely maintaining the target BLER of 10⁻². The approach significantly enhances robustness and generalization capability of link adaptation.

This work proposes integrating network coding as a forward error correction mechanism into 5G systems to replace conventional feedback-based HARQ/ARQ protocols, which suffer from high latency and inefficient resource utilization due to retransmissions. By introducing network coding at the wireless standard interface for the first time, the proposed approach significantly reduces transmission delay and enhances resource efficiency. Through rigorous mathematical modeling, network slicing simulations, and modular protocol stack design, the solution ensures in-order packet delivery while improving throughput and mitigating resource contention with other coexisting applications. The results demonstrate a promising new paradigm for future 6G protocol design that prioritizes low latency and high spectral efficiency without relying on feedback-driven retransmission mechanisms.