Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Semantic HARQ for Intelligent Transportation Systems: Joint Source-Channel Coding-Powered Reliable Retransmissions

Yongkang Li, Xu Wang, Zheng Shi, Yaru Fu

TL;DR

This paper tackles the reliability challenge of semantic communication in ITS by integrating Hybrid Automatic Repeat reQuest (HARQ) with Joint Source-Channel Coding (JSCC). It introduces a generative signal reconstructor and a local knowledge base to enable semantic error detection and correction, and proposes three sem-HARQ schemes: sem-HARQ-I, sem-HARQ-WC, and sem-HARQ-SC, each with distinct strategies for leveraging prior semantic information across retransmissions. A transformer-based semantic error detector and a local KB-based training framework underpin the system, enabling end-to-end training of the semantic encoder/decoder and the HARQ enhancements. Simulation results using BLEU and BERT-based sentence similarity demonstrate substantial gains over a No-HARQ baseline, especially at low SNR, validating the effectiveness of feature-level and synonym-based semantic combining for reliable ITS communications.

Abstract

The surge of data traffic in Intelligent Transportation Systems (ITS) places a significant challenge on limited wireless resources. Semantic communication, which transmits essential semantics of the raw data, offers a promising solution by reducing redundancy and improving spectrum efficiency. However, high vehicle mobility, dynamic channel conditions, and dense vehicular networks severely impact transmission reliability in ITS. To address these limitations, we integrate Hybrid Automatic Repeat reQuest (HARQ) with Joint Source-Channel Coding (JSCC) to provide reliable semantic communications for ITS. To counteract the adverse effects of time-varying fading channels and noise, we propose a generative signal reconstructor module supported by a local knowledge base, which employs a discriminator for channel error detection and a conditional generative network for error correction. We propose three innovative semantic HARQ (sem-HARQ) schemes, Type I sem-HARQ (sem-HARQ-I), sem-HARQ with weighted combining (sem-HARQ-WC), and sem-HARQ with synonymous combining (sem-HARQ-SC) to enable reliable JSCC-based semantic communications. At the transmitter, both sem-HARQ-I and sem-HARQ-WC retransmit the same semantic signals, while sem-HARQ-SC introduces redundant semantics across different HARQ rounds through synonymous mapping. At the receiver, sem-HARQ-I performs semantic decoding based solely on the currently received signal. In contrast, sem-HARQ-WC enhances reliability by fusing the current received semantic signal with prior erroneous signals at the feature or decision level, thereby exploiting semantic information from failed HARQ rounds. Similarly, sem-HARQ-SC employs feature-level combining, leveraging incremental semantic redundancy to merge semantic features from retransmissions.

Semantic HARQ for Intelligent Transportation Systems: Joint Source-Channel Coding-Powered Reliable Retransmissions

TL;DR

This paper tackles the reliability challenge of semantic communication in ITS by integrating Hybrid Automatic Repeat reQuest (HARQ) with Joint Source-Channel Coding (JSCC). It introduces a generative signal reconstructor and a local knowledge base to enable semantic error detection and correction, and proposes three sem-HARQ schemes: sem-HARQ-I, sem-HARQ-WC, and sem-HARQ-SC, each with distinct strategies for leveraging prior semantic information across retransmissions. A transformer-based semantic error detector and a local KB-based training framework underpin the system, enabling end-to-end training of the semantic encoder/decoder and the HARQ enhancements. Simulation results using BLEU and BERT-based sentence similarity demonstrate substantial gains over a No-HARQ baseline, especially at low SNR, validating the effectiveness of feature-level and synonym-based semantic combining for reliable ITS communications.

Abstract

The surge of data traffic in Intelligent Transportation Systems (ITS) places a significant challenge on limited wireless resources. Semantic communication, which transmits essential semantics of the raw data, offers a promising solution by reducing redundancy and improving spectrum efficiency. However, high vehicle mobility, dynamic channel conditions, and dense vehicular networks severely impact transmission reliability in ITS. To address these limitations, we integrate Hybrid Automatic Repeat reQuest (HARQ) with Joint Source-Channel Coding (JSCC) to provide reliable semantic communications for ITS. To counteract the adverse effects of time-varying fading channels and noise, we propose a generative signal reconstructor module supported by a local knowledge base, which employs a discriminator for channel error detection and a conditional generative network for error correction. We propose three innovative semantic HARQ (sem-HARQ) schemes, Type I sem-HARQ (sem-HARQ-I), sem-HARQ with weighted combining (sem-HARQ-WC), and sem-HARQ with synonymous combining (sem-HARQ-SC) to enable reliable JSCC-based semantic communications. At the transmitter, both sem-HARQ-I and sem-HARQ-WC retransmit the same semantic signals, while sem-HARQ-SC introduces redundant semantics across different HARQ rounds through synonymous mapping. At the receiver, sem-HARQ-I performs semantic decoding based solely on the currently received signal. In contrast, sem-HARQ-WC enhances reliability by fusing the current received semantic signal with prior erroneous signals at the feature or decision level, thereby exploiting semantic information from failed HARQ rounds. Similarly, sem-HARQ-SC employs feature-level combining, leveraging incremental semantic redundancy to merge semantic features from retransmissions.

Paper Structure

This paper contains 31 sections, 1 theorem, 18 equations, 16 figures.

Table of Contents

  1. Introduction
  2. Background
  3. Related Works
  4. Contributions
  5. Structure of the Paper
  6. System Model and Performance Metrics
  7. Sem-HARQ Transmission
  8. System Model
  9. Knowledge Base
  10. Semantic Communication System
  11. Reconstruction and Detector Modules for HARQ
  12. Performance Metrics of Semantic Communications
  13. BLEU Score
  14. Sentence Similarity
  15. Sem-HARQ Schemes and Semantic Combination Strategy
  16. ...and 16 more sections

Key Result

Theorem 1

Suppose the VC dimension of the hypothesis space $\mathcal{H}$ is finite and given by $d$, $m$ is the sample size, and let $h \in \mathcal{H}$. For any $m > d$ and $0 < \xi < 1$, we have where $E(h)$ and $\widehat{E}(h)$ represent the semantic distortion generalization error and the semantic distortion empirical error, respectively.

Figures (16)

  • Figure 1: The framework of proposed sem-HARQ communication system.
  • Figure 2: The framework of proposed sem-HARQ communication system.
  • Figure 3: The proposed sem-HARQ-I scheme.
  • Figure 4: sem-HARQ-WC-FC
  • Figure 5: sem-HARQ-WC-DC
  • ...and 11 more figures

Theorems & Definitions (1)

  • Theorem 1