A Survey on Intermediate Fusion Methods for Collaborative Perception Categorized by Real World Challenges

TL;DR

This survey addresses how intermediate fusion in collaborative perception balances high accuracy with practical communication constraints in autonomous driving. It categorizes existing methods by real-world challenges—transmission efficiency, localization and pose errors, communication issues, heterogeneity, adversarial threats, and domain shift—and surveys representative techniques and their evaluation metrics. The authors highlight how feature-level sharing can achieve strong detection performance under bandwidth constraints, while emphasizing gaps in real-world scalability, dataset diversity, and robust synchronization. The work provides a taxonomy to guide future research toward scalable, privacy-conscious, and robust collaborative perception systems with dynamic collaboration strategies.

Abstract

This survey analyzes intermediate fusion methods in collaborative perception for autonomous driving, categorized by real-world challenges. We examine various methods, detailing their features and the evaluation metrics they employ. The focus is on addressing challenges like transmission efficiency, localization errors, communication disruptions, and heterogeneity. Moreover, we explore strategies to counter adversarial attacks and defenses, as well as approaches to adapt to domain shifts. The objective is to present an overview of how intermediate fusion methods effectively meet these diverse challenges, highlighting their role in advancing the field of collaborative perception in autonomous driving.