Monte Carlo Localization (MCL)

I am currently investigating potential applications of Monte Carlo methods to state estimation in mobile robotics. Our new approach to mobile robot localization applies sample-based representations of the three-dimensional state space of the robot. When the robot moves or senses, sampling / importance re-sampling is applied to propagate the belief over time. Monte Carlo Localization (short MCL) has been tested extensively using different types of sensors such as sonars, laser range-finders, and cameras. The method has been shown to be extremely efficient even when globally localizing a mobile robot from scratch. Furthermore, the approach is able to robustly and accurately track a robot's position (joint work with Frank Dellaert, Wolfram Burgard, and Sebastian Thrun).

Most recently, we applied MCL to multi-robot scenarios. When one robot detects another, probabilistic detection models are used to synchronize the individual robots' believes, thereby reducing the uncertainty of both robots during localization. The constraint propagation is implemented using sample sets.  Density trees are employed to integrate information from one robot into another robot's belief.

Take  a look at some animations showing MCL in action!

Relevant publications

  • D. Fox, W. Burgard, H. Kruppa, and S. Thrun.

  • A Probabilistic Approach to Collaborative Multi-Robot Localization.
    Auonomous Robots, 8, (3), 2000.
      Slides of a talk including animations!
  • D. Fox, W. Burgard, F. Dellaert, and S. Thrun.

  • Monte carlo localization: Efficient position estimation for mobile robots.
    In Proc. of the National Conference on Artificial Intelligence (AAAI), 1999.
  •  F. Dellaert, W. Burgard, D. Fox, and S. Thrun.

  • Using the condensation algorithm for robust, vision-based mobile robot localization.
    In Proc. of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR), 1999.