Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Intelligible Protocol Learning for Resource Allocation in 6G O-RAN Slicing

Farhad Rezazadeh, Hatim Chergui, Shuaib Siddiqui, Josep Mangues, Houbing Song, Walid Saad, Mehdi Bennis

TL;DR

This work tackles inter-slice contention in 6G O-RAN slicing by introducing STEP, a multi-agent DRL framework that couples deep Q-learning with an information bottleneck to drive emergent, concise inter-agent protocols. By enforcing an IB-driven stochastic bottleneck, STEP compresses state and message information through a latent representation, controlled by a KL-divergence term and a trade-off parameter $\beta$, while enabling dual-actions: resource allocation and communication signaling. In a three-slice inter-slice conflict scenario, STEP dramatically reduces conflicts (up to 6.06× vs predefined protocols and 3.4× vs MADRL baseline), lowers latency (median down to $0$ ms from $4$ ms), and improves CPU utilization (up to $1.4×$), validating both performance gains and protocol interpretability. The approach is designed to be O-RAN compliant, with deployment as rApps/xApps in the Near-RT RIC via E2 interfaces, and it outlines future directions in communication-space design, scalability through meta-learning, and cross-layer protocol integration, highlighting significant practical impact for dynamic, interoperable 6G network slicing.

Abstract

An adaptive standardized protocol is essential for addressing inter-slice resource contention and conflict in network slicing. Traditional protocol standardization is a cumbersome task that yields hardcoded predefined protocols, resulting in increased costs and delayed rollout. Going beyond these limitations, this paper proposes a novel multi-agent deep reinforcement learning (MADRL) communication framework called standalone explainable protocol (STEP) for future sixth-generation (6G) open radio access network (O-RAN) slicing. As new conditions arise and affect network operation, resource orchestration agents adapt their communication messages to promote the emergence of a protocol on-the-fly, which enables the mitigation of conflict and resource contention between network slices. STEP weaves together the notion of information bottleneck (IB) theory with deep Q-network (DQN) learning concepts. By incorporating a stochastic bottleneck layer -- inspired by variational autoencoders (VAEs) -- STEP imposes an information-theoretic constraint for emergent inter-agent communication. This ensures that agents exchange concise and meaningful information, preventing resource waste and enhancing the overall system performance. The learned protocols enhance interpretability, laying a robust foundation for standardizing next-generation 6G networks. By considering an O-RAN compliant network slicing resource allocation problem, a conflict resolution protocol is developed. In particular, the results demonstrate that, on average, STEP reduces inter-slice conflicts by up to 6.06x compared to a predefined protocol method. Furthermore, in comparison with an MADRL baseline, STEP achieves 1.4x and 3.5x lower resource underutilization and latency, respectively.

Intelligible Protocol Learning for Resource Allocation in 6G O-RAN Slicing

TL;DR

This work tackles inter-slice contention in 6G O-RAN slicing by introducing STEP, a multi-agent DRL framework that couples deep Q-learning with an information bottleneck to drive emergent, concise inter-agent protocols. By enforcing an IB-driven stochastic bottleneck, STEP compresses state and message information through a latent representation, controlled by a KL-divergence term and a trade-off parameter , while enabling dual-actions: resource allocation and communication signaling. In a three-slice inter-slice conflict scenario, STEP dramatically reduces conflicts (up to 6.06× vs predefined protocols and 3.4× vs MADRL baseline), lowers latency (median down to ms from ms), and improves CPU utilization (up to ), validating both performance gains and protocol interpretability. The approach is designed to be O-RAN compliant, with deployment as rApps/xApps in the Near-RT RIC via E2 interfaces, and it outlines future directions in communication-space design, scalability through meta-learning, and cross-layer protocol integration, highlighting significant practical impact for dynamic, interoperable 6G network slicing.

Abstract

An adaptive standardized protocol is essential for addressing inter-slice resource contention and conflict in network slicing. Traditional protocol standardization is a cumbersome task that yields hardcoded predefined protocols, resulting in increased costs and delayed rollout. Going beyond these limitations, this paper proposes a novel multi-agent deep reinforcement learning (MADRL) communication framework called standalone explainable protocol (STEP) for future sixth-generation (6G) open radio access network (O-RAN) slicing. As new conditions arise and affect network operation, resource orchestration agents adapt their communication messages to promote the emergence of a protocol on-the-fly, which enables the mitigation of conflict and resource contention between network slices. STEP weaves together the notion of information bottleneck (IB) theory with deep Q-network (DQN) learning concepts. By incorporating a stochastic bottleneck layer -- inspired by variational autoencoders (VAEs) -- STEP imposes an information-theoretic constraint for emergent inter-agent communication. This ensures that agents exchange concise and meaningful information, preventing resource waste and enhancing the overall system performance. The learned protocols enhance interpretability, laying a robust foundation for standardizing next-generation 6G networks. By considering an O-RAN compliant network slicing resource allocation problem, a conflict resolution protocol is developed. In particular, the results demonstrate that, on average, STEP reduces inter-slice conflicts by up to 6.06x compared to a predefined protocol method. Furthermore, in comparison with an MADRL baseline, STEP achieves 1.4x and 3.5x lower resource underutilization and latency, respectively.
Paper Structure (18 sections, 6 figures)

This paper contains 18 sections, 6 figures.

Table of Contents

  1. Introduction
  2. Technical Challenges in 6G Network Slicing
  3. Agent Definition
  4. Inter-slice Decisions Conflicts
  5. Isolation vs. Monitoring Data Sharing
  6. Poor Generalization in Unseen Tasks
  7. STEP Framework
  8. Intelligible Protocol Learning
  9. O-RAN Slicing Compliance
  10. Performance Evaluation of STEP Framework
  11. Use Case: Learning Inter-Slice Conflict Resolution Protocol
  12. Network Architecture and Experiment
  13. Numerical Results
  14. Future Research Directions
  15. Communication Space Design
  16. ...and 3 more sections

Figures (6)

  • Figure 1: The STEP workflow for an individual agent involves observing a compressed version of state and communication spaces, thereby discarding redundancy and facilitating the interpretation of emerging protocols. The network slicing environment is detailed in Fig. \ref{['fig:XDRL_architecture']}.
  • Figure 2: Architecture of the inter-slice conflict resolution use case, with one agent per slice, where the O-RAN and Edge domains form the network slicing environment.
  • Figure 3: Evaluation of average computing resource allocation conflict number among three agents. All assessments were carried out on the server delineated in Section \ref{['server-spec']}.
  • Figure 4: (Up) CDF representation of average transmission and computation latencies across three slices, (Down) Comparative analysis of STEP performance regarding average resource allocation conflicts across three agents with varied communication action sizes. Unlike other experiments, this test was conducted under an exceptionally stringent conflict threshold.
  • Figure 5: A comparative view of baseline vs. STEP approaches in terms of CPU utilization performance. To provide a clearer visual representation, the curves are smoothed with respect to confidence bands and standard deviation.
  • ...and 1 more figures