从传统到基于深度学习的 SLAM 算法研究

靖添 徐(徐州医科大学，中国)
婷 杨(徐州医科大学，中国)

摘要

本文系统梳理了从传统SLAM到基于深度学习的SLAM算法的研究进展。首先介绍了SLAM算法的基本概念与系统结构，包括前端处理、后端优化、回环检测等核心模块。随后分析了传统SLAM系统如视觉SLAM和激光SLAM的特点并指出其在动态环境和低纹理的低适应性。在此基础上，重点探讨了深度学习技术如卷积神经网络（CNN）、Transformer模型、循环神经网络（RNN）等在SLAM技术中特征提取、位姿估计及语义理解等方面的融合应用。总结了LIFT-SLAM、DeepVo、DLD-SLAM、NeRF-SLAM等代表性算法的优势与不足。最后指出深度学习技术可显著提升SLAM系统在复杂环境下的适应性，为SLAM技术的进一步发展提供参考。

关键词

SLAM；深度学习；深度神经网络

参考

Bruno H M S,Colombini E L.LIFT-SLAM:a deep-learning feature-based monocular visual SLAM method[J].Neurocomputing,2021,455;97-100.

Wang S,Clark R,Wen H,et al.DeepVO:Towards End-to-End Visual Odometry with Deep Recurrent Convolutional Neural Networks[J]2017 IEEE International Conference on Robotics and Automation(ICRA),Singapore,2017,2043-2050.

Zhu B,Yu A,Hou B,et al.A Novel Visual SLAM Based on Multiple Deep Neural Networks[J].Applied Sciences,2023,13(17):9630

Zhou D,Luo Y,Zhang Q,et al.A Lightweight Neural Network for Loop Closure Detection in Indoor Visual SLAM[J].International Journal of Computational Intelligence Systems,2023,16(1):49.

Yu H,Wang Q,Yan C,et al.DLD-SLAM:RGB-D Visual Simultaneous Localisation and Mapping in Indoor Dynamic Environments Based on Deep Learning[J].Remote Sensing,2024,16(2):246.

Sun Y,Hu J,Yun J,et al.Multi-Objective Location and Mapping Based on Deep Learning and Visual Slam[J].Sensors,2022,22(19):7576.

Ao S,Hu Q,Yang B,et al.SpinNet:Learning a General Surface Descriptor for 3D Point Cloud Registration[Z],//2021IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR),Nashville,TN,USA,2021:11748-11757.

Li Y,Harada T.Lepard:Learning partial point cloud matching in rigid and deformable scenes[Z].//2022 IEEE/CVF Conference onComputerVision and PatternRecognition(CVPR),NewOrleans,LA,USA,2022:5544-5554.

Chen X,Läbe T,Milioto A,et al.OverlapNet:A siamese network for computing LiDAR scan similarity with applications to loop closing and Localization[J].Autonomous Robots,2021,46(1):61–81.

喻伟东,鲁静,程晗蕾.基于NeRF的SLAM研究综述[J].计算机系统应用,2025,34(04):18-33.

Yang S,Scherer S.CubeSLAM:Monocular 3-D Object SLAM[J].IEEE Transactions on Robotics,2019,35(4):925–938.

Nicholson L,Milford M,Sünderhauf N.QuadricSLAM:Dual Quadrics from Object Detections as Landmarks in Object-oriented SLAM[Z].IEEE Robotics and Automation Letters,2019,4(1):1-8.

DOI: http://dx.doi.org/10.12345/bdai.v6i5.32066