Problem
Research questions and friction points this paper is trying to address.
collaborative perception
sensor corruptions
communication corruptions
semantic recovery
multi-agent systems
Diff-KD: Diffusion-based Knowledge Distillation for Collaborative Perception under Corruptions
🤖 AI Summary
This work addresses the vulnerability of multi-agent cooperative perception in real-world scenarios to sensor and communication disturbances, which often degrade semantic information in ways that existing methods cannot actively recover. To tackle this challenge, the study introduces conditional diffusion mechanisms into cooperative perception for the first time, proposing a teacher–student framework with progressive knowledge distillation to reconstruct corrupted local features. Furthermore, an adaptive gating fusion module is designed to dynamically weight neighboring agents’ information based on the ego vehicle’s perceptual reliability. Evaluated on the OPV2V and DAIR-V2X datasets under seven distinct interference types, the proposed approach achieves state-of-the-art detection accuracy and calibration robustness, significantly enhancing the overall system resilience.
📝 Abstract
Multi-agent collaborative perception enables autonomous systems to overcome individual sensing limits through collective intelligence. However, real-world sensor and communication corruptions severely undermine this advantage. Crucially, existing approaches treat corruptions as static perturbations or passively conform to corrupted inputs, failing to actively recover the underlying clean semantics. To address this limitation, we introduce Diff-KD, a framework that integrates diffusion-based generative refinement into teacher-student knowledge distillation for robust collaborative perception. Diff-KD features two core components: (i) Progressive Knowledge Distillation (PKD), which treats local feature restoration as a conditional diffusion process to recover global semantics from corrupted observations; and (ii) Adaptive Gated Fusion (AGF), which dynamically weights neighbors based on ego reliability during fusion. Evaluated on OPV2V and DAIR-V2X under seven corruption types, Diff-KD achieves state-of-the-art performance in both detection accuracy and calibration robustness.
Innovation
Methods, ideas, or system contributions that make the work stand out.
diffusion-based knowledge distillation
collaborative perception
corruption robustness
progressive knowledge distillation
adaptive gated fusion
🔎 Similar Papers
No similar papers found.
