Ilmu Komputer & AI editorial
Open AccessOA2026
Adaptive Peer Clustering with Hierarchical Random Linear Network Coding for Resilient Decentralized Wireless Networks
APC-RLNC: EWMA-driven clustering and multi-tier RLNC with O(√T) regret for AI-native 6G collectives
Navaneetha Krishnan Kamalakannan; Harinisri Velmurugan· 2026· DOI 10.48550/arXiv.2608.26040
The core problem
Decentralized wireless collectives—vehicular swarms, IoT clusters, and edge AI networks—demand communication protocols that remain robust under dynamic topologies and heterogeneous link quality. Random Linear Network Coding (RLNC) offers algebraic resilience against packet erasures, but its performance degrades when peers experience diverse channel conditions. This paper introduces Adaptive Peer Clustering with Hierarchical RLNC (APC-RLNC), a system that dynamically groups peers by exponentially weighted moving average (EWMA) reliability metrics and applies multi-tier network coding within and across clusters. The work formalizes the clustering optimization problem, derives closed-form decoding probability bounds for Markov erasure channels, and proves regret for online reconfiguration under the Follow-the-Regularized-Leader (FTRL) framework. The authors position adaptive clustering as a foundational primitive for AI-native 6G wireless systems.
Innovation
Evaluation across high-mobility vehicular networks, burst-error channels, and adversarial interference shows consistent gains over state-of-the-art baselines. APC-RLNC achieves 5.2–9.8 percentage-point improvements in packet delivery ratio (PDR), 10–23% latency reductions, and up to 30% higher node retention. The system scales linearly to 500+ nodes while maintaining real-time reconfiguration overhead below 3%. These results are observed across diverse scenarios, including burst-error channels and adversarial interference, demonstrating robustness under dynamic conditions.
Decentralized wireless collectives—vehicular swarms, IoT clusters, and edge AI networks—demand communication protocols that remain robust under dynamic topologies and heterogeneous link quality. Random Linear Network Coding (RLNC) offers algebraic resilience against packet erasures, but its performance degrades when peers experience diverse channel conditions. This paper introduces Adaptive Peer Clustering with Hierarchical RLNC (APC-RLNC), a system that dynamically groups peers by exponentially weighted moving average (EWMA) reliability metrics and applies multi-tier network coding within and across clusters. The work formalizes the clustering optimization problem, derives closed-form decoding probability bounds for Markov erasure channels, and proves regret for online reconfiguration under the Follow-the-Regularized-Leader (FTRL) framework. The authors position adaptive clustering as a foundational primitive for AI-native 6G wireless systems.
APC-RLNC operates in three stages: (1) reliability estimation via EWMA of link quality, (2) adaptive peer clustering, and (3) hierarchical RLNC encoding/decoding. Each peer maintains an EWMA reliability score updated as:
Why it matters
The performance gains stem from APC-RLNC's ability to isolate unreliable peers into separate clusters, preventing them from degrading the coding efficiency of reliable clusters. The hierarchical RLNC design ensures that erasures within a cluster are compensated by intra-cluster coding, while inter-cluster coding provides resilience against cluster-level failures. The regret bound guarantees that online reconfiguration converges quickly even under non-stationary channel conditions. The linear scalability to 500+ nodes and sub-3% overhead make APC-RLNC suitable for real-time edge deployments. The authors argue that adaptive clustering is a foundational primitive for AI-native 6G wireless systems, where heterogeneous devices and dynamic topologies are the norm. Limitations include the need for a reliable feedback channel for EWMA updates and the assumption of Markov erasure channels; future work could extend to more general channel models and integrate with AI-driven predictive scheduling.
Who should read this
CS practitioners and researchers
Opening member content…