collab_localization

A repository of simulation and experimental work in Decentralized Collaborative Localization

MatLab Package

The MatLab package provided in this repository contains a vectorized and parallelized simulation framework for testing collaborative localization algorithms. This framework can handle an indeterminate number of vehicles, sensors, and models. Additionally, Monte-Carlo analysis can be quickly produced

Experimental Setup

Experiments can easily be setup using a input *.m file such as this. Simply definiting initial conditions for each vehicle and the desired control inputs is all one needs to run the simulations.

nCars = 2;
in_name = 'par';

x0 = [ ...
    2,0,0,2;...
   -2,0,0,2]';

acc_cmd = [ ...
    0,  0,  0;...
    4,  0,  0];

del_cmd = [ ...
    0,      0,      0;...
    4,      0,      0;...
    8,      0,      0;...
    12,     0,      0];

Monte-Carlo Analysis

Using very large numbers of numbers of simulations, it is possible to create error distributions for various localization algorithms. In order to take advantage of a multi-core CPU in addition to simultaneous, vectorized simultions, simply run

multithread.m 

Otherwise, for single-threaded (but still vectorized) simulations, run

main.m

ROS Pacakge

The Included ROS node sets up and runs the collaborative localization algorithm used in experimentation with the UWB sensors and the autonomous trolley.

Dependencies

Currently, running the experimental ROS node relies on the following packages

• ubxtranslator
• decawave_ros
• vn300

Launch

Using ROS, the following launch file (ROS/launch) initializes

roslaunch collab_localization estimation.launch