Overview

Place recognition is an essential component to address the problem of robot localization and loop closure in visual Simultane- ously Localization and Mapping (SLAM). Long-term navigation of robots in outdoor environments introduces new challenges to enable lifelong SLAM, including the strong appearance change resulting from vegetation, weather, and illumination variations across various times of the day, different days, months, or even seasons. Several examples of the same place with significant appearance variations are illustrated in the following figure.

SRAL

We propose a novel Shared Representative Appearance Learning (SRAL) to address long-term place recognition. Different from previous techniques based on a single feature modality or concatenation of multiple features, our SRAL approach autonomously learns shared features that are representative in all scene scenarios, then fuses the features together to represent the long-term appearance of environments observed by a robot during lifelong navigation. By formulating SRAL as a regularized optimization problem, we use structured sparsity-inducing norms to model interrelationships of feature modalities. In addition, an optimization algorithm is developed to efficiently solve the formulated optimization problem, which holds a theoretical convergence guarantee.


Code and Citation

Download the Matlab code for our SRAL approach HERE. The code provided here is for non-commercial research/educational use only.

Please cite the following paper if you use our code or method. Detailed information of the SRAL approach is also available in the paper.

  @INPROCEEDINGS{han2016life,
      AUTHOR    = {Fei Han AND Xue Yang AND Yiming Deng AND Mark Rentschler AND Dejun Yang AND Hao Zhang},
      TITLE     = {SRAL: Shared Representative Appearance Learning for Long-Term Visual Place Recognition},
      BOOKTITLE = {IEEE International Conference on Robotics and Automation (ICRA)},
      YEAR      = {Submitted}
  }

or

  @INPROCEEDINGS{han2016life,
      AUTHOR    = {Fei Han AND Xue Yang AND Yiming Deng AND Mark Rentschler AND Dejun Yang AND Hao Zhang},
      TITLE     = {Life-Long Place Recognition by Shared Representative Appearance Learning},
      BOOKTITLE = {Workshop on Robotics: Science and Systems},
      YEAR      = {2016}
  }

Questions and Suggestions

Please contact Fei Han: fhan -AT- mines -DOT- edu