Plug-and-Play Reweighting for Resilient Collaborative Decision-Making in Connected Autonomous Driving

Jul 16, 2026

Upload by Noah

Collaborative decision-making is crucial for multi-robot systems such as connected autonomous vehicles, yet it is fragile under perceptual noise and adversarial attacks from collaborators. Existing defenses often require attack-specific retraining or strong perturbation assumptions, limiting practicality. This work introduces a Resilient Collaborative Decision-Making (RCDM) framework with attention-based encoding and decoding, plus a plug-and-play reweighting module that down-weights corrupted inputs using neighborhood consistency and deviation from the local median. High-fidelity simulations with noise and five attack types show up to 26% improvement and state-of-the-art resilient performance.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages8

Reading Time20 min

Introduction

Vulnerabilities in collaborative decision-making

Existing resilience approaches

Plug-and-Play Reweighting for Resilient Collaborative Decision-Making in Connected Autonomous Driving Jiewen Liu 1 , Rui Liu2 , Matthew Lee3 , Ming C. Lin2 , Xiaorui Liu 1 , and Peng Gao 1 arXiv :2607 . 10037v1 [ cs .RO] 10 Jul 2026 Abstract—Collaborative decision-making is a fundamental capability in multi-robot systems, such as connected autonomous vehicles. However, perceptual noise and adversarial attacks in collaborators can severely affect decision reliability. Overall, existing methods typically rely on retraining with attack-specific defenses or on restrictive perturbation assumptions to improve resilience, which limits their practicality. In this paper, we propose a novel Resilient Collaborative DecisionMaking (RCDM) framework that consists of an attention-based encoder for extracting individual robot perceptual embeddingsand an attention-based decoder for fusing collaborator perceptions and making decisions. To improve resilience to corrupted observations, we design a novel plug-and-play reweighting module that down-weights the influence of corrupted inputs by analyzing the consistency of neighborhood points relative to the local structure and assigning smaller weights to points that deviate strongly from the local median. This module can be seamlessly integrated into attention-based collaborative decision-making without requiring additional training. We evaluate our method in high-fidelity simulations, considering perceptual noise and five types of attacks across diverse accidentprone scenarios. Experimental results demonstrate that our approach consistently outperforms existing methods by up to 26% and achieves state-of-the-art resilient performance. I. INTRODUCTION Multi-robot systems have been widely studied for decades due to their scalability, reliability and parallelism. To enable efficient multi-robot collaboration, a fundamental capability is collaborative decision-making, with the goal of enabling the robots to make informed decisions by leveraging knowledge shared and integrated across teammates. It has a variety of applications, such as multi-robot collaborative search and rescue [1], [2], connected autonomous driving [3], and collaborative manufacturing [4] . However, collaborative decision-making is highly vulnerable to corrupted observations, which may arise from perceptual noise in individual robots or from adversarial data transmitted by collaborators [5] . Such corrupted observations distort the collective understanding of the environment and can ultimately lead to unsafe decisions. As shown in Figure 1, the yellow collaborator assists the ego vehicle by sharing its perception of an oncoming red vehicle that may 1Jiewen Liu, Xiaorui Liu and Peng Gao are with the Department of Computer Science, North Carolina State University. Email: {jliu222, xliu96, [pgao5](pgao5}@ncsu.edu.2 Rui Liu)[}](pgao5}@ncsu.edu.2 Rui Liu)[@ncsu.edu.](pgao5}@ncsu.edu.2 Rui Liu)[2](pgao5}@ncsu.edu.2 Rui Liu)[ Rui Liu](pgao5}@ncsu.edu.2 Rui Liu) and Ming C. Lin are with the Department of Computer Science, University of Maryland, College Park. Email: {ruiliu, [lin](lin}@umd.edu.3 Matthew Lee is with University of North Carolina at Chapel)[}](lin}@umd.edu.3 Matthew Lee is with University of North Carolina at Chapel)[@umd.edu.](lin}@umd.edu.3 Matthew Lee is with University of North Carolina at Chapel)[3](lin}@umd.edu.3 Matthew Lee is with University of North Carolina at Chapel)[ Matthew Lee is with University of North Carolina at Chapel](lin}@umd.edu.3 Matthew Lee is with University of North Carolina at Chapel)[ ](lin}@umd.edu.3 Matthew Lee is with University of North Carolina at Chapel)[Hill. Email: matthewlee01234@gmail.com](Hill. Email: matthewlee01234@gmail.com). © 2026 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new

FAQs

What problem does the paper address in connected autonomous driving?

It addresses how collaborative decision-making can fail when collaborators provide corrupted observations due to perceptual noise or adversarial attacks, leading to unreliable and unsafe decisions.

How does the proposed RCDM framework combine collaborators’ information?

It uses an attention-based encoder to extract each robot’s perceptual embeddings and an attention-based decoder to fuse collaborator perceptions and produce the final decision.

What is the plug-and-play reweighting module and how does it work?

The module reduces the influence of corrupted inputs by analyzing the consistency of neighborhood points with the local structure and assigning smaller weights to points that deviate strongly from the local median, without requiring additional training.

Jul 16, 2026

Upload by Noah