PaperlayerView exact revision on arXiv

Source-linked AI summary

Goal-Oriented and Semantic Communication in 6G AI-Native Networks: The 6G-GOALS Approach

Emilio Calvanese Strinati, Paolo Di Lorenzo, Vincenzo Sciancalepore, Adnan Aijaz, Marios Kountouris, Deniz Gündüz, Petar Popovski, Mohamed Sana, Photios A. Stavrou, Beatriz Soret, Nicola Cordeschi, Simone Scardapane, Mattia Merluzzi, Lanfranco Zanzi, Mauro Boldi Renato, Tony Quek, Nicola di Pietro, Olivier Forceville, Francesca Costanzo, Peizheng Li

arXiv:2402.07573v1eess.SP

TL;DR

6G-GOALS addresses the challenge of interconnecting AI modules sustainably when conventional transmission may ignore information’s relevance to receiver objectives. It proposes an O-RAN-based semantic and goal-oriented communication framework, and presents its project vision alongside two proof-of-concept use cases.

  • Problem

    6G must interconnect AI modules sustainably, while conventional transmission can ignore information’s relevance to receiver objectives.

    Source: §I · p. 1

  • Method

    6G-GOALS proposes an O-RAN-based AI/ML architecture and semantic communication pillars, including causal semantic representation and reasoning for goal-oriented communication.

    Source: §I · p. 2, §II · p. 2, §III · p. 3

  • Results

    The paper presents the project vision and describes two proof-of-concept use cases: hardware semantic-aware communication and semantic-aware collaborative robots.

    Source: §IV · p. 5, §V · p. 6

  • Takeaways & Limitations

    The project seeks practical, standardizable 6G applications and resource allocation targeting energy and spectrum efficiency, reduced EMF exposure, and legacy interoperability.

    Source: §I · p. 2, §V · p. 6

Abstract

from arXiv · show

Recent advances in AI technologies have notably expanded device intelligence, fostering federation and cooperation among distributed AI agents. These advancements impose new requirements on future 6G mobile network architectures. To meet these demands, it is essential to transcend classical boundaries and integrate communication, computation, control, and intelligence. This paper presents the 6G-GOALS approach to goal-oriented and semantic communications for AI-Native 6G Networks. The proposed approach incorporates semantic, pragmatic, and goal-oriented communication into AI-native technologies, aiming to facilitate information exchange between intelligent agents in a more relevant, effective, and timely manner, thereby optimizing bandwidth, latency, energy, and electromagnetic field (EMF) radiation. The focus is on distilling data to its most relevant form and terse representation, aligning with the source's intent or the destination's objectives and context, or serving a specific goal. 6G-GOALS builds on three fundamental pillars: i) AI-enhanced semantic data representation, sensing, compression, and communication, ii) foundational AI reasoning and causal semantic data representation, contextual relevance, and value for goal-oriented effectiveness, and iii) sustainability enabled by more efficient wireless services. Finally, we illustrate two proof-of-concepts implementing semantic, goal-oriented, and pragmatic communication principles in near-future use cases. Our study covers the project's vision, methodologies, and potential impact.

I. INTRODUCTION

As 6G shifts toward natively interconnecting AI modules, 6G-GOALS proposes communication that prioritizes information relevant to receiver objectives over content-blind transmission. The project lays groundwork for semantic and goal-oriented communications through an AI/ML-based architecture, semantic reasoning, and resource allocation intended to support efficient, interoperable services.

I. INTRODUCTION

  • 6G networks are envisioned to interconnect AI modules sustainably through convergence of communication, computation, control, and learning.The paper frames this as a new phase beyond earlier approaches focused on throughput, coverage, and connectivity.Source: §I · p. 1
  • 6G-GOALS challenges content-blind transmission and targets information with practical significance for receiver learning or inference objectives.Its vision is to transmit only the most pertinent information for those objectives.Source: §I · p. 1
  • The project proposes an AI/ML-based architecture atop hierarchical O-RAN, with a semantic plane, semantic RICs, and semantic reasoning engines.It also investigates distributed reasoning and goal-oriented actuation in time-varying data scenarios.Source: §I · p. 2
  • Building on initial results,, the project investigates potential effects and countermeasures when distinct semantic settings cause communication errors at semantic and effectiveness levels, potentially inducing interpretation errors.It incorporates causal inference in representation learning for efficient semantic representations of unstructured data and robustness to possible distribution shifts.Source: §I · p. 2
  • Proposed resource-allocation strategies target energy and spectrum efficiency, reduced EMF exposure, and interoperability with legacy semantic-agnostic agents.The project also seeks backward compatibility with data-oriented communication systems.Source: §I · p. 2

II. 6G-GOALS SYSTEM ARCHITECTURE

6G-GOALS builds its semantic and goal-oriented system architecture on O-RAN. Its design combines intelligent network functions and a semantic plane with scalable RAN operation and coexistence with legacy components.

II. 6G-GOALS SYSTEM ARCHITECTURE

  • The architecture builds on O-RAN and uses its programmability, disaggregation, openness, and RAN intelligent controllers.It considers a disaggregated RAN in which the base station is split into RU, DU, and CU.Source: §II · pp. 2–3
  • The design targets an AI-native 6G system for semantic and goal-oriented communications, with intelligent functions for semantic extraction, representation, and resource optimization.A semantic plane is intended to support semantic services, knowledge-driven reasoning, user experience, and system efficiency.Source: §II · p. 2
  • The architecture aims to make the RAN adaptable for large-scale semantic and goal-oriented communications.Source: §II · p. 2
  • New semantic-aware components are intended to operate alongside existing legacy system components.Source: §II · p. 2

6G-GOALS Architectural Blocks

The architectural blocks place semantic service management, applications, and processing across the network, from the SMO layer and RIC to RAN components, edge devices, and the core. They also include a shared knowledge base and support for generative AI.

6G-GOALS Architectural Blocks

  • The semantic engine at the SMO layer orchestrates semantic-information processing, model lifecycle management, and user-experience management.Generative AI fine-tuned for vertical applications can be deployed in the engine; pruned generative models may be deployed at RU/DU level.Source: 6G-GOALS Architectural Blocks · p. 3
  • The Semantic RIC provides a programmable, extensible platform for semantic xApps, connected to CU control and user planes and the DU.It supports service requirements from non-real-time through near-real-time to real-time.Source: 6G-GOALS Architectural Blocks · p. 3
  • The RAN semantic plane spans CU, DU, and RU, coordinating semantic information flow and enabling extraction, encoding, and JSCC in semantic-aware DU/RU components.The application plane is orthogonal to the semantic plane and interfaces semantic applications across edge devices or user equipment.Source: 6G-GOALS Architectural Blocks · p. 3
  • UE and edge devices process and interpret raw images, video, and sensor inputs through semantic modules that extract and contextualize information.Source: 6G-GOALS Architectural Blocks · p. 3
  • The knowledge base supports semantic processing, including semantic-model development, training, validation, decoding, and interpretation.Source: 6G-GOALS Architectural Blocks · p. 3
  • The enhanced core network can support semantic RAN and management functions, including AI-optimization metrics via NWDAF and policy or QoS exposure via NEF.Source: 6G-GOALS Architectural Blocks · p. 3

III. RESEARCH ON INNOVATIVE TECHNOLOGY ENABLERS

The 6G-GOALS framework aims to move beyond sense-compute-connect-control models toward semantic communication-based AI architectures, protocols, and services. Its research agenda is organized around semantic data processing, timing-aware communication, and sustainability.

III. RESEARCH ON INNOVATIVE TECHNOLOGY ENABLERS

  • One pillar develops AI-empowered semantic data representation, sensing, compression, and communication.Source: §III · p. 3
  • A second pillar focuses on timing-aware semantic communication for distributed reasoning and actuation.Source: §III · p. 3
  • A third pillar addresses 6G sustainability through a semantic-empowered RAN.Source: §III · p. 3

A. AI-empowered semantic data representation, sensing, compression, and communication

The first pillar develops foundations for semantic and goal-oriented communication by reducing transmitted data while optimizing desired performance under complexity and physical constraints.

A. AI-empowered semantic data representation, sensing, compression, and communication

  • The first pillar develops methods for data acquisition, extraction, representation, compression, and caching to reduce transmitted data while optimizing desired performance under complexity and physical constraints.The methodologies are detailed in the sequel.Source: A. AI-empowered semantic data… · p. 3

1) Mathematical Definitions and Fundamental Limits:

The project studies information-theoretic fundamentals and trade-offs in semantic communication, then develops task-relevant data representations and sensing under energy, accuracy, and latency constraints.

1) Mathematical Definitions and Fundamental Limits:

  • The project examines semantic information representation, network trade-offs, semantic metrics, and value of information in networked systems.Its scope includes multiuser sampling and compression, convex and non-convex optimization, and strategies for energy savings and complexity reduction.Source: 1) Mathematical Definitions and… · pp. 3–4
  • The project also explores generative AI, Kolmogorov complexity, and topological models to balance representation richness with semantic information extraction.These are among its strategies for improving cost-performance trade-offs and reducing energy use and complexity.Source: 1) Mathematical Definitions and… · p. 4

2) Semantic Data Acquisition, Representation, and Compression:

  • For task-specific relevance, the project develops semantic data acquisition, representation, and compression methods under energy, accuracy, and latency constraints.Approaches include manifold learning, topological neural networks, optimal representations, rate-limited-channel algorithms, and methods addressing the rate-distortion-perception trade-off.Source: 2) Semantic Data Acquisition,… · p. 4

3) Semantic Source and Channel Coding Schemes:

The project advances semantic source and channel coding, timing-aware distributed reasoning, and methods for recovering meaning when communicating agents use different languages or models.

3) Semantic Source and Channel Coding Schemes:

  • DeepJSCC research targets image and video transmission, seeking greater complexity, training, and generalizability performance than current convolutional models and practical next-generation mobile-network solutions.Planned work includes modular, standards-compatible architectures, transformer-aided media transmission, topological-DNN semantic coding, adaptable schemes for different signal sources, and wireless DNN-model delivery.Source: 3) Semantic Source and Channel Coding… · p. 4

B. Timing-aware semantic communication for distributed reasoning and actuation

  • Distributed reasoning and actuation work develops AI/ML-aided semantic communication and reasoning algorithms that respect timing constraints for coordination and control under 6G-GOALS KPIs.The research area focuses on timing-varying data scenarios.Source: B. Timing-aware semantic communication… · p. 4

1) Foundations of Reasoning through Semantic-Empowered Communications:

  • Different languages and models can create semantic noise and performance issues,, which 6G-GOALS addresses through domain adaptation for sender-receiver mismatches.Its methodology uses optimal transport for semantic recovery linked to domain adaptation and rate-distortion theory, plus JSCC schemes robust to model mismatches.Source: 1) Foundations of Reasoning through… · p. 4

2) Semantic Communications under Timing Constraints:

6G-GOALS pursues semantic communication under timing constraints, causal representations for pragmatic communication, and adaptive RAN mechanisms to balance sustainability-related performance trade-offs.

2) Semantic Communications under Timing Constraints:

  • The project links information relevance and value to time-aware metrics, with AI/ML methods for transmission, decision-making, and information exchange under timing constraints.It also proposes stochastic optimization, reinforcement learning, context-specific coding, and prioritized communication strategies.Source: 2) Semantic Communications under… · p. 4

3) Causal Semantic Representations and Pragmatic Communications:

  • To address distribution shifts that challenge statistical ML models, the project develops causal semantic representations and distributed causal reinforcement learning for communication and action strategies.Methods include causal representation learning from time-varying and unstructured data, multi-view learning, sequential optimization, scalable causal estimation, and low-delay compression for semantic reconstruction.Source: 3) Causal Semantic Representations and… · p. 4

C. 6G sustainability via semantic-empowered RAN

  • Semantic-performance metrics will guide adaptive RAN mechanisms in optimizing trade-offs among spectrum and energy efficiency, reliability, latency, and end-to-end communication, learning, and control performance.The pillar aims to connect architectural, theoretical, and algorithmic solutions with operationally adaptive mechanisms.Source: C. 6G sustainability via… · p. 4

1) Semantic control strategies:

6G-GOALS develops semantic control and resource-management strategies for semantic-oriented communications, aiming to improve efficiency while balancing semantic and traditional communication.

1) Semantic control strategies:

  • Semantic control targets efficient two-way L1/L2 signaling, real-time O-RAN control, and compatibility with existing vendor-specific systems.The project also proposes implicit AI-based control signaling and metrics for error handling and message reformulation.Source: 1) Semantic control strategies · pp. 4–5
  • Fig. 2 presents the experimental setting for a proof of concept on real-time semantic communication.Source: Figure 2 · p. 5

2) Connect-Compute Network Resource Management for Semantic Communications:

  • Resource management prioritizes goal-effectiveness and semantics while balancing semantic and traditional communications through AI-driven spectrum and energy management.The approach also explores minimizing transmitted data while preserving communication goals and using Over-the-Air Computing for functions across multiple transmissions.Source: 2) Connect-Compute Network Resource… · p. 5

IV. PROOFS OF CONCEPT

The proof-of-concept program compares semantic and goal-oriented approaches with conventional communications; its hardware implementation studies real-time model transmission over a constrained wireless link.

IV. PROOFS OF CONCEPT

  • The proof-of-concept program includes hardware semantic communication and semantic-aware collaborative robotics, comparing their performance with conventional schemes.Source: §IV · p. 5

A. Hardware implementation of semantic communication

  • JetsonNano terminals train and run inference, while software-defined radios map model parameters to channel symbols under varied channel and mobility conditions.The point-to-point PoC aims to let a receiver perform inference on new samples using a transmitted model.Source: A. Hardware implementation of semantic… · p. 5
  • The PoC considers sending samples for receiver-side training or sending a transmitter-trained model; bandwidth, latency, and channel state constrain transmissions.Its primary focus is joint source-channel coding for model transmission, with noise injection during training and distillation intended to improve robustness to channel noise.Source: A. Hardware implementation of semantic… · p. 5

B. Semantic Communications-enabled Cooperative Robotics

The cooperative-robotics PoC combines semantic-aware sensing and networking to compare against raw command-and-feedback communication. The conclusion identifies architecture, communication limits, resource efficiency, compatibility, and benchmarking as challenges for practical 6G applications.

B. Semantic Communications-enabled Cooperative Robotics

  • The robotics trial compares semantic communications with raw command-and-feedback messages in collaborative warehousing and logistics tasks.Robots process heterogeneous sensor data in real time using GPU capabilities; the testbed includes a semantic-aware physical layer and knowledge base for semantic codecs.Source: B. Semantic Communications-enabled… · p. 5
  • A semantic-aware controller manages models and timely responses, while O-RU processing, virtualized software, and new QoS classes support semantic communication and task allocation.The network setup includes a 3GPP-compliant core network and Open RAN system with semantic awareness and an enhanced RIC platform.Source: B. Semantic Communications-enabled… · pp. 5–6

V. CONCLUSIONS

  • The project identifies four challenges: a semantic plane and modules, communication limits, sustainable resource use with legacy compatibility, and experimental benchmarks.Source: §V · p. 6

Related papers

Loading 2402.07573v1…