Objective-Aware Policy Orchestration for O-RAN Service Provisioning

The gist The proposed objective-aware policy orchestration framework selects the most appropriate provisioning policy for each control window based on the current network context and operator objective.

Objective-Aware Policy Orchestration

This framework proposes an objective-aware policy orchestration framework that selects the most appropriate provisioning policy for each control window based on the current network context and operator objective The framework orchestrates a portfolio of heterogeneous provisioning experts using contextual-bandit learning Two policy selectors, Linear Thompson Sampling and a Bootstrap Tree ensemble, are evaluated against static provisioning policies and a deterministic rule-based orchestrator Experimental results show that learning-based policy orchestration can effectively adapt policy selection to varying operating conditions, improving overall objective-aware performance while preserving the modularity of existing provisioning algorithms.

Framework Components and Context

The orchestration layer operates at a slower timescale and can be conceptually placed in the Non-Real Time (RT) RAN Intelligent Controller (RIC) or the Service Management and Orchestration (SMO) framework. The orchestration decision depends on both the operator objective and the network context. The context includes the performance of the completed observation window, available compute and transport resources, congestion indicators, failure information, a target-window traffic forecast, and operational information such as the currently active policy. The operator objective indicates whether the next decision should prioritize Grade of Service (GoS), energy consumption, latency, congestion mitigation, or a balanced combination.

Provisioning Expert Portfolio

The orchestrator selects one policy map from the expert portfolio introduced in Section III. The portfolio contains four heterogeneous provisioning strategies: Shortest-Path (SP), Minimum Latency (MinLat), Minimum Energy (MinEnergy), and PPO-based learned provisioning. Each expert implements a complete provisioning policy capable of embedding and routing all service-chain requests during one control window. The orchestrator does not modify these experts or their internal optimization procedures Instead, it learns under which operating conditions each expert is expected to provide the highest performance score.

Contextual Policy Selection

The policy-selection problem is formulated as a contextual-bandit task. Given the current context xt, the selector estimates the expected score of every eligible policy map Rˆ(axt), and selects at = arg max a∈At Rˆ(axt). Two complementary contextual-bandit models are investigated. The first learner, Linear Thompson (LT) Sampling, provides regularized actionspecific reward models and explicit uncertainty estimates The second learner, a Bootstrap Tree Ensemble (BTE), captures nonlinear interactions between objectives, traffic, failures, and resource conditions.

Objective-Aware Outcome Evaluation

The proposed framework is objective-aware, meaning that the preferred provisioning policy depends on the operational objective specified by the network operator. The primary score Sprim equals vG, vE, vL, or vu for GoS, energy, latency, or congestion objectives respectively. For balanced operation uses w⊤B v where v = (vG, vE, vL, vu, vc, vf). For compliant outcomes Rt = [0.75Sprim + 0.25Usec −Csw − Crec] where Csw = 0.03δt and Crec = 0.02δt, and otherwise Rt = −1 + 0.9fcons.

Performance Evaluation

The BTE selector achieved the highest objective-aware score while simultaneously obtaining the highest GoS. These results indicate that learning to select among heterogeneous provisioning experts can provide measurable improvements over relying on a single static provisioning policy. The comparison with the static provisioning policies further illustrates the motivation for policy-level orchestration. The nonlinear tree ensemble is better able to capture conditional interactions between operator objectives, traffic conditions, and network state.

Future Work

Future work will focus on extending the current framework in several directions. First, the present implementation selects a single policy map for the entire network; future work will investigate service-specific and region-specific policy orchestration. Second, the current evaluation is based on canonical one-step counterfactual execution, whereas future work will consider fully closed-loop orchestration in which each selected policy influences subsequent network states. Finally, the proposed contextual-bandit selector will be integrated into a complete agentic orchestration framework in which large language models translate operator intent into structured objectives while the learning component remains responsible for objective-aware policy selection.

How it works

The proposed architecture separates the orchestration process into four logical stages: i) observation of the current network state, ii) selection of the provisioning policy, iii) execution of the selected provisioning expert, iv) and objectiveaware evaluation of the resulting network performance. At the beginning of every control window, the orchestrator observes the current network conditions together with the operator objective and selects one provisioning policy map from a predefined portfolio. The selected policy remains active during the entire target window, while all request-level provisioning decisions are delegated to the chosen expert.

Provisioning Expert Portfolio

The portfolio contains four heterogeneous provisioning strategies: Shortest-Path (SP), Minimum Latency (MinLat), Minimum Energy (MinEnergy), and PPO-based learned provisioning. The orchestrator does not modify these experts or their internal optimization procedures Instead, it learns under which operating conditions each expert is expected to provide the highest performance score.

Contextual Policy Selection

The policy-selection problem is formulated as a contextual-bandit task. Given the current context xt, the selector estimates the expected score of every eligible policy map Rˆ(axt), and selects at = arg max a∈At Rˆ(axt). Two complementary contextual-bandit models are investigated.

Objective-Aware Outcome Evaluation

The proposed framework is objective-aware, meaning that the preferred provisioning policy depends on the operational objective specified by the network operator. The primary score Sprim equals vG, vE, vL, or vu for GoS, energy, latency, or congestion objectives respectively. For balanced operation uses w⊤B v where v = (vG, vE, vL, vu, vc, vf).

Deterministic Rule-Based Orchestrator

As a deterministic baseline, a rule-based orchestrator was implemented using a fixed sequence of operational checks and expert-selection rules. If no safety mechanism is triggered, the decision is driven by the operator objective Latency-oriented objectives select the minimum-latency expert, energy-oriented objectives select the minimum-energy expert, while all remaining situations default to the MinLat provisioning policy.

Results Summary

Among all evaluated methods, the BTE selector achieved the highest objective-aware score while simultaneously obtaining the highest GoS. This indicates that learning to select among heterogeneous provisioning experts can provide measurable improvements over relying on a single static provisioning policy. The deterministic rule-based orchestrator achieves competitive energy consumption while maintaining acceptable GoS.

1. textbfObjective-Aware Policy Selection via Contextual Bandits (BTE/LT): The improved system can dynamically select one approved provisioning policy map for each control window based on the current network context and operator objective, enabling flexible adaptation without modifying the underlying experts.

2. textbfHandling Heterogeneous Expert Strengths: The BTE selector learns operating regions in which different provisioning strategies become preferable, allowing it to exploit the strengths of multiple experts where a single static policy fails, as demonstrated by achieving the highest objective-aware score while simultaneously obtaining the highest GoS.

3. textbfIncreased Adaptability Beyond Static Policies: The system moves beyond fixed objectives by evaluating outcomes using an objective-aware score defined in Section IV-E, which incorporates primary and secondary performance preferences, allowing it to adapt selection based on whether the operator objective is GoS, energy efficiency, latency, congestion, or a balanced combination.

4. textbfImproved Decision Making Under Uncertainty: The use of two complementary learners—Linear Thompson Sampling (LT) and Bootstrap Tree Ensemble (BTE)—allows the system to quantify uncertainty; for instance, LT provides explicit uncertainty estimates, while BTE uses prediction variability across individual trees as an uncertainty estimate for Thompson-style action selection.

5. textbf Separation of Concerns: The framework clearly separates high-level orchestration from low-level execution by ensuring the orchestrator selects one approved provisioning policy map for the next control window and that the lower-level provisioning expert remains responsible for processing individual servicechain requests.

O-RAN enables the deployment of heterogeneous control and provisioning algorithms through open interfaces and programmable intelligent controllers. However, most existing intelligent O-RAN solutions rely on a single optimization strategy designed for a fixed operational objective, limiting their ability to adapt when network conditions or operator priorities change. This paper proposes an objective-aware policy orchestration framework that selects the most appropriate provisioning policy for each control window based on the current network context and operator objective. The framework orchestrates a portfolio of heterogeneous provisioning experts using contextual-bandit learning. Two policy selectors, Linear Thompson Sampling and a Bootstrap Tree ensemble, are evaluated against static provisioning policies and a deterministic rule-based orchestrator. Experimental results show that learning-based policy orchestration can effectively adapt policy selection to varying operating conditions, improving overall objective-aware performance while preserving the modularity of existing provisioning algorithms.