Conflict-Aware Autonomous Resource Orchestration for 5G Edge Slicing Using Multi-Agent Systems and LLMs

Nathan R. Prediger; Rodrigo Moreira; Juliano Araujo Wickboldt; Tereza C. Carvalho; Pedro Oliveira; Paulo Novais; Flávio de Oliveira Silva

arXiv · 2610.14903

Conflict-Aware Autonomous Resource Orchestration for 5G Edge Slicing Using Multi-Agent Systems and LLMs

Nathan R. Prediger·Rodrigo Moreira·Juliano Araujo Wickboldt·Tereza C. Carvalho·Pedro Oliveira·Paulo Novais·Flávio de Oliveira Silva

View original source ↗Read source document ↗Export metadata ↓

Abstract

5G edge network slicing must reconcile two coupled requirements: understanding high-level operator intent and reallocating scarce computing and bandwidth resources under dynamic contention. Centralized optimization can face computational and responsiveness challenges under rapidly changing edge contention, and Multi-Agent System (MAS)-based auctions can support resource arbitration but still depend on operator-defined economic parameters and priorities. Embedding a Large Language Model (LLM) directly in a recurring control loop, by contrast, may introduce non-negligible inference latency and non-deterministic behavior. This paper proposes a two-timescale architecture that addresses this gap: an LLM-based Strategist translates natural-language network intents into economic policy parameters (budget, income, and bidding-aggressiveness) at the macro scale, while a deterministic, Vickrey-inspired second-price mechanism run by SPADE-based slice agents resolves recurring resource contention at the operational scale, with the resulting allocations enforced by patching Kubernetes CPU limits and Calico CNI bandwidth annotations directly on User Plane Function pods. We evaluate the architecture end-to-end on a real Kubernetes/Open5GS edge cluster across two complementary scenarios, each repeated 30 times: a Smart-City scenario, in which natural-language public-safety intents propagate into measurable changes in slice bids and bandwidth allocations within a few auction rounds, and a Hierarchical-slicing scenario, in which the mechanism provides priority-dependent resource and QoE differentiation, adapts to abrupt demand changes, and preserves configured minimum guarantees while maintaining near-complete allocation of the shared bandwidth pool.

Explore related subjects

cs.NI cs.DC cs.MA

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

Nathan R. Prediger, Rodrigo Moreira, Juliano Araujo Wickboldt, Tereza C. Carvalho, Pedro Oliveira, Paulo Novais, Flávio de Oliveira Silva. 2026-10-10. Conflict-Aware Autonomous Resource Orchestration for 5G Edge Slicing Using Multi-Agent Systems and LLMs. https://arxiv.org/abs/2610.14903

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

Energy Consumption in Next Generation Radio Access Networks

Historically, in macro-cell-based networks, the radio access network (RAN) portion accounts for the largest share of energy consumption in mobile networks. With greater densification, the transition to small cells and increased virtualization of the radio processing functions, the relative energy use of the different network segments is changing. At the same time, open radio access networks (O-RAN) have emerged as a promising approach to support this evolution through open interfaces that enable a multi-vendor environment, support for hierarchical intelligent controls, and simplified, cost-effective radio units that facilitate large-scale deployments. This paper examines the impact of these changes on energy consumption in next-generation RAN architectures through transaction-based energy models. Transaction-based network energy models were developed in 2009 to understand the evolution of energy use in networks. Here we update these models with today's hardware and apply recently developed models for emerging architectures, considering current access point densification trends. Results indicate that efficiency improvements have followed predicted trends, which were largely based on semiconductor Moore's Law-related evolution. Emerging RAN architectures show improved energy per bit, although the processing energy becomes the largest contributor. Densification increases total energy consumption, but the relative contributions of processing and transport energy depend strongly on the location of the baseband processing.

cs.NI↗

arXiv2026

When Unlearning Fails: Reliable Data Deletion under Post-Training in Agent Networks

Embedded AI agents on edge devices can learn as a federation and keep improving after deployment. Each agent collects new trajectories with the current policy, and the edge server aggregates them in later rounds. In such self-improving agent networks, deleting data is harder than a one-time model repair. The data to be deleted may have already shaped later retained trajectories, so retraining or model-side unlearning leaves an influence echo that returns as the network keeps learning. We propose Muting Unlearned Trajectories' Echoes (MUTE) for reliable deletion over edge networks. MUTE traces the echo from a lightweight ledger at the edge server, without moving raw trajectories off the devices. It erases the current residue with a forget-retain update, contains high-influence retained trajectories, and audits later behavior to schedule further erasure under the uplink budget and the response deadline. On LIBERO, MUTE keeps the regenerated share of forgotten behavior at or below 0.401 for every request type, against at least 0.466 for five recent methods.

cs.NI↗

arXiv2026

QoE-Aware Service Function Chain Orchestration in Space-Air-Ground Integrated Networks for In-Flight Internet

In remote regions lacking terrestrial infrastructure, aviation users increasingly demand Internet services comparable to ground cellular networks. This paper proposes a space-air-ground integrated network (SAGIN) architecture that combines ground networks, civil aviation networks, and low Earth orbit (LEO) satellites. Based on real North Atlantic flight data, a heterogeneous aviation Internet is constructed to support diverse aircraft service requirements. A composite Quality of Experience (QoE) metric, characterized by transmission delay, jitter, and content similarity, guides efficient resource utilization. The dynamic connectivity of SAGIN is modeled using a reconfigurable time-expanded graph (RTEG). With Network Function Virtualization (NFV), data delivery is formulated as a Service Function Chain (SFC) deployment across ground stations, relay aircraft, and satellites, resulting in a long-term QoE maximization problem. A tailored Multi-Agent Deep Deterministic Policy Gradient (MADDPG) algorithm incorporating LCSS-based path filtering and softmax-based action selection is developed. Simulation results demonstrate that the proposed approach enables seamless connectivity for heterogeneous aircraft and consistently outperforms game-theoretic and MATD3 methods in terms of QoE.

cs.NI↗