From Automated to Autonomous: Hierarchical Agent-native Network Architecture (HANA)

Binghan Wu, Shoufeng Wang, Yunxin Liu, Ya-Qin Zhang, Joseph Sifakis, Ye Ouyang

arXiv:2605.20608 · 2026-05-22 공개 · arXiv · PDF

multi-agent-systems throughput-optimization hierarchical-architecture fault-recovery agent-native-intelligence deliberative-governance autonomous-networks self-aware-agents

Abstract

Realizing Level 4/5 Autonomous Networks (AN) demands a shift from static automation to agent-native intelligence. Current operations, reliant on rigid scripts, lack the cognitive agency to handle off-nominal conditions. To address this, this letter proposes a hierarchical multi-agent reference architecture enabling high-level autonomy. The framework features a Dual-Driven Orchestrator that coordinates specialized Executive Agents, supported by a shared Public Memory for unified domain knowledge. A key innovation is the integration of agent self-awareness, which empowers the system to harmonize deliberative strategic governance with reflexive fault recovery. We instantiate and validate this architecture within a 5G Core environment. Case studies demonstrate that the system sustains critical throughput under congestion and reduces Mean Time to Repair (MTTR) by 86%, confirming its efficacy in unifying strategic planning with operational resilience.

한국어 요약

한 줄 요약

**[자율 네트워크 / Multi-Agent Architecture]** HANA는 Dual-Driven Orchestrator·Executive Agent·Public Memory의 hierarchical 멀티에이전트 reference architecture로 L4/5 autonomy 지향, 5G Core에서 임계 throughput 유지·MTTR 86% 감소 검증.

핵심 기여도

핵심 아이디어

L4/5 autonomous network는 rigid script로 도달 불가능하며, hierarchical 멀티에이전트 architecture에서 dual-driven orchestrator가 specialized agent를 조정하고 self-awareness로 deliberative governance와 reflexive recovery를 통합해야 high autonomy를 달성한다.

기술적 접근법

주요 결과

의의 및 한계

**의의**: L4/5 자율 네트워크의 구체적 architectural blueprint, dual-driven orchestrator·public memory·self-awareness의 명확한 역할 분리, 5G Core 실증으로 실용성 확보, MTTR 86% 감소의 실질적 운영 가치. **한계**: 5G Core 단일 환경 검증으로 다른 네트워크 stack 일반화 추가 검증, agent self-awareness 메커니즘 구체 구현 abstract에서 미상세, 매우 dynamic 변화에 대한 확장성 제한 가능.

실용적 활용