Robust Monte Carlo Localization for Mobile Robots Thomas Coffee Based on: Thrun S, Fox D, Burgard W, Dellaert F Robust Monte Carlo Localization for Mobile Robots (2001) Artificial Intelligence 128(1-2): 99-141 Image: Thrun et al. 2001
The Problem of Localization “Using sensory information to locate the robot in its environment is the most fundamental problem to providing a mobile robot with autonomous capabilities.” (Cox, 1991) Image: Fox et al. 1999
Tracking vs. Global Localization
Global Localization Requires Multi-Modal Belief Representations Image: Fox et al. 1999
Global Localization for a Mobile Robot Image: Thrun et al. 2001
Multi-Hypothesis Kalman Filtering Image: Roumeliotis et al. 2000
Real Errors are Non-Gaussian! Image: Thrun et al. 2001
Markov Localization (ML) Image: Fox et al. 1999
Particle Filters to the Rescue! Image: Thrun et al. 2001
Monte Carlo Localization (MCL) Image: Thrun et al. 2001
Monte Carlo Localization (MCL) Image: Thrun et al. 2001
Monte Carlo Localization (MCL) Image: Thrun et al. 2001
How MCL Works
Performance of MCL vs. ML Image: Thrun et al. 2001
Simulated Object Localization with MCL Image: Thrun et al. 2001
Better Sensors = Larger Errors? Image: Thrun et al. 2001
Object Localization Failure with MCL Image: Thrun et al. 2001
What Went Wrong?
A Quick Fix for MCL Image: Thrun et al. 2001
Key Idea: Dual Sampling MCL
Kernel Density Trees: Computing Densities from Particle Fields Recursive sampling above threshold Calculate densities by sum of weights in leaf divided by volume of leaf Equivalent to maximum likelihood estimation of piecewise constant density functions Like particle filters, concentrates effort in most useful regions Image: Fox et al. 2000
Results of Dual MCL Image: Thrun et al. 2001
Mixture-MCL: Best of Both Breeds Image: Thrun et al. 2001
Results for Small Samples Image: Thrun et al. 2001
Results for the Kidnapping Problem Image: Thrun et al. 2001
Real Implementation of Mixture-MCL: Sampling Poses from Observations
Mixture-MCL in Action Image: Thrun et al. 2001
Mixture-MCL in Action Image: Thrun et al. 2001
Mixture-MCL in Action Image: Thrun et al. 2001
Results for Real Implementation Image: Thrun et al. 2001
Is Mixture-MCL Efficient? Image: Thrun et al. 2001
Almost as Fast as Standard MCL! Image: Thrun et al. 2001
Advantages of Mixture-MCL
Related Work and Applications
Limitations and Assumptions
Future Extensions to Mixture-MCL
Thank you! Image: Thrun et al. 1999