无人系统视觉SLAM技术发展现状简析

A survey of visual SLAM in unmanned systems

视觉SLAM(visual SLAM,vSLAM)已成为无人系统在未知环境中实现全自主导航和环境感知不可或缺的重要组成部分.鉴于此,简要介绍vSLAM的发展历程和典型构成,总结两种前端位姿估计方法(特征法与直接法)和两种后端优化方法(非线性滤波与非线性优化)的典型代表及优缺点.在此基础上,依次按照耦合方式和后端优化方法的不同对视觉/惯性SLAM的相关研究成果进行分类.进一步分析代表性开源vSLAM框架的异同,给出公开数据集下的性能对比,并分别从泛化能力、高级感知能力、动态适应性和多传感器集成化等方面对当前vSLAM所面临的挑战进行阐述,探讨vSLAM未来的发展趋势和方向,为研究者提供有益参考.

Visual simultaneous localization and mapping(vSLAM)plays a vital role in fully autonomous navigation and perception of unmanned systems in unknown environment.Following a brief introduction to the history and typical structure of vSLAM,a summary about state-of-the-art representatives,advantages and disadvantages of two front-end pose estimation methods(feature based vs.photometric based)and two back-end optimization methods(nonlinear filter vs.nonlinear optimization)is given.On this basis,relevant achievements of visual inertial SLAM(VI-SLAM)are classified into different categories according to their coupled types and back-end optimization methods.Furthermore,the similarities and differences of representative open-source vSLAM frameworks are compared and analyzed,alongside with the comparison of their performance under public datasets.Finally,challenges faced by current vSLAM are elaborated from the aspects of scenario generalization,advanced perception,dynamic adaptability and multi-sensor integration.Future development trends and directions of vSLAM are also discussed to serve as an useful guide for researchers.