Orion uses memory to improve 6g network slice placement decisions
ORION: An Outcome-Driven LLM Advisor for RL-Based Multi-Domain 6G Slice Placement
Machine LearningArtificial IntelligenceMultiagent SystemsNetworking and Internet Architecture
Summary
Placing virtual parts of 6G networks across different company networks is a very tricky problem, especially when requests arrive all the time and must be handled quickly. The paper shows that no single fixed method works best in all situations, and many good solutions go unnoticed. To address this, the authors built ORION, a system that uses language models and reinforcement learning agents to help decide how to split and place network slices. ORION remembers which plans worked or failed in real network conditions and uses this memory to improve future decisions, leading to more requests being accepted.
What this means in practice
- •For network operators: Improve decisions on where to place 6G network slices across different administrative domains to increase accepted requests.
- •For telecom infrastructure engineers: Use ORION’s outcome-driven planning layer to optimize virtual network function placements for unpredictable 6G slice demands.
Authors
Mahmoud S. Mayaleh, Fetia Bannour, Nour-El-Houda Yellas
Abstract
Deploying 6G network slices across administrative domains requires an online decision for each arriving request: which domain hosts each virtual network function, how traffic crosses domain boundaries, and whether the slice is admitted. This joint decision problem is NP-hard under continuous arrivals. Experiments on a parametrized family of multi-domain topologies show that no fixed planning strategy dominates across all regimes: the performance ranking reverses between topology classes, and feasible placements exist for most rejected requests but remain undiscovered. These findings identify the plan layer as the decisive performance lever. We introduce ORION, a hierarchical orchestrator in which two LLM agents translate service intents into structured requests and generate abstract placement plans. A cooperative multi-agent RL coordinator selects the cross-domain partition, treating the plan as a persistent soft prior, while per-domain agents place the functions. ORION main contribution lies at the plan layer: an outcome-driven episodic memory that records, for each completed attempt, the network-condition profile, the proposed plan, and its verified feasibility outcome. The memory feeds verified successes and labelled failures back to the planner, thereby closing the adaptation loop using deployment feasibility rather than a scalar reward assigned to the planner's own proposals. To isolate the effect of the plan quality, we evaluate the planner on held-out topology families while keeping the downstream selection and execution mechanisms fixed. Under this controlled setting, the proposed memory improves admission by $4.6$--$11.5$ percentage points over the same planner without memory, while importance-based retention accounts for most of the improvement over a first-in-first-out store.