Autonomous Networks in 3GPP

Autonomous networks are mobile networks that monitor, analyse, decide, and act with little or no human involvement. This page explains how 3GPP defines network autonomy — the levels, the control loop, the building blocks like SON and analytics — and how it all fits together toward self-running 6G networks.

Status: Autonomous-network capabilities are normative in 3GPP (the 28-series), built up across Releases 15–20 and led by the management group (SA5). Full autonomy is a direction of travel, not a finished destination.

Levels of autonomy

3GPP defines a grading scale — Levels of Autonomous Network — so a network's degree of self-management can be measured, much like the levels used for self-driving cars:

L0 — No automation: humans do everything.
L1 — Assisted: the system recommends, humans decide and act.
L2 — Partial: the system executes within a narrow, predefined scope.
L3 — Conditional: humans set objectives and handle exceptions; the system manages within them.
L4 — High: humans set the overall intent; the system handles operations in its domain.
L5 — Full: the system runs autonomously across all scenarios, no human needed at run time.

Levels are applied per domain — the radio side might be at one level while the core is at another.

The control loop at the heart of it

Autonomy is built on a closed control loop — a repeating cycle of:

Monitor — collect events, status, and performance data from the network.
Analyse — find the cause of issues, predict problems, and recommend actions.
Decide — choose what to do (a human decides in an "open" loop; the system decides in a "closed" loop).
Execute — apply the change, then feed the results back into the next cycle.

3GPP turns the closed loop itself into a managed object, so loops can be created, configured, monitored, and even cascaded (one loop's output feeding another).

The building blocks

Several specifications combine to make autonomy work:

Self-Organising Networks (SON) — the classic automation pillars: self-configuration (auto-setup of new equipment), self-optimisation (continuous tuning of coverage, mobility, load, energy), and self-healing (detecting and recovering from failures).
Management Data Analytics (MDA) — the intelligence layer that turns raw data into insight: what happened, why, what will happen next, and what to do about it.
AI/ML management — governs the machine-learning models used in management: how they are trained, tested, deployed, run, and retrained, including distributed (federated) learning.
Intent-driven management — the outcome-based interface where operators state goals and the system pursues them (see the companion article on intent management).
Service assurance — the closed loop focused specifically on keeping communication-service quality at the promised level.

How the pieces relate

Think of it as a stack:

Intent sits on top — it states the desired outcome.
Closed control loops execute autonomously to reach it.
MDA provides the analysis that drives decisions, optionally using AI/ML (governed by the AI/ML management spec).
SON functions are often the hands that carry out the changes.
Underneath, performance measurements and KPIs let the system judge its own success.

Intent and SON are complementary: intent is the outcome-level ask, SON is one of the function-level mechanisms used to deliver it.

AI as an enabler, not a given

3GPP is deliberate that AI/ML is optional inside analytics — a management function may use statistics, rules, or machine learning. When it does use ML, a dedicated framework governs the model lifecycle so the automation stays trustworthy and auditable.

Why it matters for 6G

6G networks will be larger, denser, and more dynamic than anything before — spanning terrestrial and non-terrestrial access, AI-native functions, and tight service requirements. That is beyond efficient hand-operation, so autonomy (higher levels driven by intent and closed loops) is central to running 6G at all.

Reference points (3GPP)

TS 28.100 — levels of autonomous network.
TS 28.567 — management of closed control loops.
TS 28.104 — management data analytics (MDA).
TS 28.105 — AI/ML management.
TS 28.312 — intent-driven management.
TS 28.313 — SON for 5G.

In short

3GPP defines network autonomy as graded levels (L0–L5), powered by closed control loops and built from SON, analytics, AI/ML management, intent, and assurance — the foundation for networks that increasingly run themselves, which 6G will depend on.