Domain-Generalized Adaptive Semantic Communication for Collaborative Perception

2026-07-27

AI Summary

This work addresses semantic degradation in vehicular-to-everything (V2X) cooperative perception caused by the coupling of observation domain shifts and unknown wireless channel impairments. To tackle this challenge, the paper proposes the RSTA framework, which, for the first time, decouples domain shift and channel distortion through theoretical decomposition of task robustness and designs corresponding algorithmic components. Prior to deployment, a semantic encoder is pretrained via cross-domain prototype alignment and cross-channel gradient consistency. At inference, only a lightweight decoder adapter—updating merely 0.21% of parameters—is fine-tuned, combined with reliability-gated entropy minimization to achieve source-free domain generalization without requiring agent synchronization. Evaluated on four V2X benchmarks, the method improves AP@0.7 by 7.2 and 5.5 points on cross-weather and cross-dataset tasks, respectively.

Abstract

We propose RSTA, a domain-generalized semantic communication framework enabling source-free V2X collaborative perception under both observation-domain shift and unseen wireless channel conditions. In V2X, received semantic tokens suffer coupled degradation from pre-transmission domain drift and in-transit channel corruption; existing methods address only one source, leaving adaptation misled by tokens that are simultaneously off-domain and physically degraded. RSTA trains a pre-deployment semantic encoder for transmission stability via cross-domain prototype alignment and cross-channel gradient consistency, and updates a lightweight in-deployment decoder adapter through reliability-gated entropy minimization that restricts gradients to tokens ranked high in both semantic relevance and channel fidelity. A theoretical task robustness decomposition links each loss term to a distinct degradation source, grounding each algorithmic component in a measurable error mode. Trained on AWGN and tested on unseen Rayleigh fading, RSTA achieves +7.2 AP@0.7 over pre-deployment domain generalization on cross-weather tasks and +5.5 on cross-dataset tasks across four V2X benchmarks, updating only 0.21\% of parameters in-deployment with zero inter-agent synchronization overhead.

Problem

• domain shift
• wireless channel corruption
• semantic communication
• collaborative perception
• V2X

Innovation

• domain generalization
• semantic communication
• V2X collaborative perception
• channel-robust adaptation
• prototype alignment