Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Beyond Target-Level: ISAC-Enabled Event-Level Sensing for Behavioral Intention Prediction

Haotian Liu, Zhiqing Wei, Yucong Du, Jiachen Wei, Xingwang Li, Zhiyong Feng

TL;DR

This work tackles event-level sensing for behavioral intention prediction in autonomous driving by leveraging Integrated Sensing and Communication (ISAC). It introduces the ASI-BIP framework, which separately processes high-rate TV and low-rate UV information via dual Bi-LSTM branches and fuses them with a Transformer to capture inter-vehicle interactions, all while handling asynchronous data through specialized signal processing. The approach demonstrates robust performance in NLoS and adverse weather, achieving an $11.4\%$ improvement in macro F1-score over sensor-based baselines and showing promising generalization to unseen intentions. Overall, the study substantiates ISAC as a viable foundation for high-level perception and decision-support in intelligent transportation systems and related industrial contexts.

Abstract

Integrated Sensing and Communication (ISAC) holds great promise for enabling event-level sensing, such as behavioral intention prediction (BIP) in autonomous driving, particularly under non-line-of-sight (NLoS) or adverse weather conditions where conventional sensors degrade. However, as a key instance of event-level sensing, ISAC-based BIP remains unexplored. To address this gap, we propose an ISAC-enabled BIP framework and validate its feasibility and effectiveness through extensive simulations. Our framework achieves robust performance in safety-critical scenarios, improving the F1-score by 11.4% over sensor-based baselines in adverse weather, thereby demonstrating ISAC's potential for intelligent event-level sensing.

Beyond Target-Level: ISAC-Enabled Event-Level Sensing for Behavioral Intention Prediction

TL;DR

This work tackles event-level sensing for behavioral intention prediction in autonomous driving by leveraging Integrated Sensing and Communication (ISAC). It introduces the ASI-BIP framework, which separately processes high-rate TV and low-rate UV information via dual Bi-LSTM branches and fuses them with a Transformer to capture inter-vehicle interactions, all while handling asynchronous data through specialized signal processing. The approach demonstrates robust performance in NLoS and adverse weather, achieving an improvement in macro F1-score over sensor-based baselines and showing promising generalization to unseen intentions. Overall, the study substantiates ISAC as a viable foundation for high-level perception and decision-support in intelligent transportation systems and related industrial contexts.

Abstract

Integrated Sensing and Communication (ISAC) holds great promise for enabling event-level sensing, such as behavioral intention prediction (BIP) in autonomous driving, particularly under non-line-of-sight (NLoS) or adverse weather conditions where conventional sensors degrade. However, as a key instance of event-level sensing, ISAC-based BIP remains unexplored. To address this gap, we propose an ISAC-enabled BIP framework and validate its feasibility and effectiveness through extensive simulations. Our framework achieves robust performance in safety-critical scenarios, improving the F1-score by 11.4% over sensor-based baselines in adverse weather, thereby demonstrating ISAC's potential for intelligent event-level sensing.
Paper Structure (13 sections, 7 equations, 5 figures, 1 table)

This paper contains 13 sections, 7 equations, 5 figures, 1 table.

Table of Contents

  1. Introduction
  2. System Model
  3. ISAC Sensing Signal Model
  4. Proposed framework
  5. BIP Problem Formulation
  6. ISAC Echo Signal Processing
  7. Proposed ASI-BIP Network
  8. Simulation Results and Analysis
  9. Simulation Setup
  10. Feasibility of the Proposed Framework
  11. Generalization ability of the Proposed Framework
  12. Superiority of the Proposed Framework
  13. Conclusion

Figures (5)

  • Figure 1: (Modified) The illustration for the BIP based on two ISAC-BSs
  • Figure 2: (Modified) Proposed ASI-BIP network architecture.
  • Figure 3: Loss and confusion matrix of the proposed framework.
  • Figure 4: Generalization ability of the proposed framework for unknown behavioral intention
  • Figure 5: The F1-scores of various behavioral intentions with different SNR