Mixmcl

2D MCL with mixture proposal distributions

Requirement

Nuklei

This library is used to perform kernel density estimation and kdtree sampling.

$ sudo apt-get install gcc g++ make pkg-config python libgsl0-dev libblas-dev liblapack-dev libboost-all-dev
$ git clone https://github.com/chungying/nuklei.git
$ cd nuklei
$ ./scons.py
$ sudo ./scons.py install

FLANN

This library is for kdtree and 1-nearest-neighbor search.

$ sudo apt-get install -y libflann-dev

ROS packages

Most packages could be installed via rosdep or apt-get. Following two packages should be installed as well.

amcl_modified
stamped_std_msgs

Functions

MCL with mixture proposals by using pre-built sampling
MCL with mixture proposals by using Evolutionary Markov Chain Monte Carlo sampling

Performance comparison

Global Localisation Problem

Some results for AMCL, MCL, MIXMCL, and MCMCL (the proposed method) are shown here.

mp: maximum particle number

ri: resample intervel

chungying/mixmcl