EUVIMEDEUVIMED entdecken
Crossref

A Survey of Deep Learning-Based Multi-Sensor Fusion for 3D Object Detection in Autonomous Driving

Yining Liu

Journal of Intelligence and Knowledge Engineering · 2026-06 · 4 (2) · 134-141

DOI: 10.62517/jike.202604219 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Intelligence and Knowledge Engineering, 2026-06. Verfasst von Yining Liu. Die Arbeit nennt 25 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Multi-sensor fusion has become a fundamental framework for robust 3D object detection in autonomous driving, as it can mitigate the inherent limitations of single-modal perception under challenging real-world conditions. This paper presents a systematic review of mainstream cross-modal fusion paradigms (early, mid-level, and late fusion) tailored to 3D detection tasks, with a focus on their core mechanisms and deployment implications. We elaborate on the critical challenges of coordinate alignment and shared representation interfaces, highlighting the Bird’s-Eye View (BEV) as a promising unified fusion space that balances spatial awareness and planning compatibility. A comprehensive comparison is conducted among three dominant 3D representations-point-based, voxel-based, and BEV-based-revealing their inherent trade-offs in computational cost, memory consumption, spatial structure preservation, and deployment efficiency. Furthermore, we categorize typical deployment-facing failure modes, including domain shift, calibration drift, sensor dropout, and adversarial attacks, and summarize corresponding layered mitigation strategies such as health monitoring, consistency checking, data augmentation, and graceful degradation. This work provides a structured overview of the state-of-the-art in multi-sensor fusion for 3D detection, offering practical insights for both academic research and industrial deployment.

Zitieren

APA

Liu, Y. (2026). A Survey of Deep Learning-Based Multi-Sensor Fusion for 3D Object Detection in Autonomous Driving. Journal of Intelligence and Knowledge Engineering, 4(2), 134-141. https://doi.org/10.62517/jike.202604219

Vancouver

Liu Y. A Survey of Deep Learning-Based Multi-Sensor Fusion for 3D Object Detection in Autonomous Driving. Journal of Intelligence and Knowledge Engineering. 2026;4(2):134-141. doi:10.62517/jike.202604219

BibTeX
@article{Liu2026,
  author = {Liu, Yining},
  title = {A Survey of Deep Learning-Based Multi-Sensor Fusion for 3D Object Detection in Autonomous Driving},
  journal = {Journal of Intelligence and Knowledge Engineering},
  year = {2026},
  volume = {4},
  number = {2},
  pages = {134--141},
  doi = {10.62517/jike.202604219}
}
Kostenlos starten →