ParCon: Noise-Robust Collaborative Perception via Multi-module Parallel Connection

TL;DR

A novel collaborative perception architecture, called ParCon, which connects multiple modules in parallel, as opposed to the sequential connections used in most other collaborative perception methods, which achieves state-of-the-art accuracy and is computationally efficient.

Abstract

In this paper, we investigate improving the perception performance of autonomous vehicles through communication with other vehicles and road infrastructures. To this end, we introduce a novel collaborative perception architecture, called ParCon, which connects multiple modules in parallel, as opposed to the sequential connections used in most other collaborative perception methods. Through extensive experiments, we demonstrate that ParCon inherits the advantages of parallel connection. Specifically, ParCon is robust to noise, as the parallel architecture allows each module to manage noise independently and complement the limitations of other modules. As a result, ParCon achieves state-of-the-art accuracy, particularly in noisy environments, such as real-world datasets, increasing detection accuracy by 6.91%. Additionally, ParCon is computationally efficient, reducing floating-point operations (FLOPs) by 11.46%.

Figures (7)

• Figure 1: About this paper. (a) and (b) present two architectures. (c) and (d) ParCon holds state-of-the-art performance with a reduced number of model parameters and under various noises (details in Section \ref{['Quantitative evaluation']})
• Figure 2: Overview of our proposed collaborative perception system. Our model consists of five steps: metadata sharing, feature extraction, feature sharing, fusion module, and detection head. The details of each component are discussed in Section \ref{['method']}.
• Figure 3: Sub-modules in ParCon. (a) Agent-wise Attention (A-Att) sub-module with HRPE. (b) Spatial-wise Attention (S-Att) sub-module. (c) Spatial-wise Convolution (S-Conv) sub-module. These modules are discussed in Section \ref{['fusion module']}.
• Figure 4: Robustness in various noise ranges on V2XSet, OPV2V, and DAIR-V2X.
• Figure 5: Robustness comparison between ParCon and ParCon-S in various noise ranges on V2XSet. ParCon-S is the sequential architecture model using the same sub-modules of ParCon.