Coop-WD: Cooperative Perception with Weighting and Denoising for Robust V2V Communication

TL;DR

This paper addresses the vulnerability of cooperative perception to V2V channel impairments by introducing Coop-WD, a hierarchical framework that couples self-supervised adaptive weighting at the vehicle level with pixel-level denoising via a conditional diffusion model. The approach explicitly handles non-stationary and time-varying distortions through a dual-stage enhancement: filtering distorted shared features and reconstructing degraded pixel-level maps, with an efficient variant Coop-WD-eco that disables denoising when distortion is severe to save computation. Extensive experiments on realistic channels (Rician, WINNER II, non-stationary V2V) and the V2V4Real dataset show that Coop-WD consistently outperforms baselines, while Coop-WD-eco significantly reduces runtime (up to 50% or more) with minimal performance loss under severe distortions. The work advances robust cooperative perception for autonomous driving by integrating generative denoising with adaptive feature weighting under realistic communication constraints, enabling more reliable multi-vehicle perception in dynamic environments.

Abstract

Cooperative perception, leveraging shared information from multiple vehicles via vehicle-to-vehicle (V2V) communication, plays a vital role in autonomous driving to alleviate the limitation of single-vehicle perception. Existing works have explored the effects of V2V communication impairments on perception precision, but they lack generalization to different levels of impairments. In this work, we propose a joint weighting and denoising framework, Coop-WD, to enhance cooperative perception subject to V2V channel impairments. In this framework, the self-supervised contrastive model and the conditional diffusion probabilistic model are adopted hierarchically for vehicle-level and pixel-level feature enhancement. An efficient variant model, Coop-WD-eco, is proposed to selectively deactivate denoising to reduce processing overhead. Rician fading, non-stationarity, and time-varying distortion are considered. Simulation results demonstrate that the proposed Coop-WD outperforms conventional benchmarks in all types of channels. Qualitative analysis with visual examples further proves the superiority of our proposed method. The proposed Coop-WD-eco achieves up to 50% reduction in computational cost under severe distortion while maintaining comparable accuracy as channel conditions improve.

Figures (7)

• Figure 1: Cooperative perception via V2V communication.
• Figure 2: System architecture of cooperative perception with V2V communication.
• Figure 3: Pixel-level diffusion and denoising processes for feature maps in cooperative perception system.
• Figure 4: The proposed joint weighting and denoising algorithm pipeline.
• Figure 5: Performance under different path loss factors and time-varying disturbances. Disturbances are simulated with a fixed time duration following Gaussian distribution. (a) Path loss factor $n$. (b) Time-varying noise levels ($\sigma_\text{SNR}$). (c) Time-varying CSI errors($\sigma_\text{CSI}$).