/dock_scan_template_matching

Primary LanguageC++

Robot Charge Station Detection using Template Matching

Overview

The dock_scan_template_matching package is a specialized ROS package. It implements algorithm for robust robot charge station detection for accurate and precise docking purpose using 2D laser scan data. Firstly downsamples model point cloud to decrease computational complexity. After that, sampled model point cloud is transformed to every point from laser scan data on 2D. After transformation of model point cloud, score is calculated according to corresponding points between model and laser scan data by comparing their distance. Bestscore gives best transformation for model point cloud. It means that best tranformation gives accurate and precise location of robot charge station for docking purpose. I am sharing my result on record of rviz visualizer;
https://youtu.be/Tps68zlMCuk

Dependencies

  • ROS (tested on Melodic)
  • PCL (tested on 1.2)
  • Eigen3
  • C++11 or later

Installation

git clone https://github.com/sefakurtipek/dock_scan_template_matching.git
catkin_make // to make build

Usage

rosbag play scanMatchBag1.bag

The user needs to write the following commands on linux terminal to take parameters and run ROS package

    cd dock_scan_template_matching/launch
    roslaunch dock_scan_template_matching params.launch

In order to visualize I used rviz. I need to remap tf and tf_static according to name of the robot. For example, I took "/r300311695" name from 'rostopic list' command

rosrun rviz rviz /tf:=/r300311695/tf /tf_static=/r300311695/tf_static

Parameters you need to know

The user may need to tune given parameters located in params.launch file.
initialGuess_yaw_degrees: initial angle guess on radian
modelSampleNum: Number of points taken from model point cloud. Higher number of model samples may increase computational complexity of the algorithm
max_distance_between_corresponding_points: Maximum distance threshold to match between model samples and laser scan data points on 2D
scoreThreshold: If bestScore is higher than this scoreThreshold, program prints BAD SCORE.
vShapeModelName: robot charge station model point cloud file. Type of file is .pcd
laserFrameName: Frame name of laser scan data
mapFrameName: Map frame name. It is used to get frame transformation.
laserScanName: Laser scan data topic name. It is used to subscribe laser scan data\
absolute_path: Absolute path of project files. It is used to give model point cloud file location.\

    <param name="initialGuess_yaw_degrees" type="double" value="265.364" />
    <param name="modelSampleNum" type="int" value="10" />
    <param name="max_distance_between_corresponding_points" type="double" value=" 0.1" />
    <param name="scoreThreshold" type="double" value="0.01" />
    <param name="vShapeModelName" type="string" value="RMRS004.pcd" />
    <param name="laserFrameName" type="string" value="/r300311695/base_front_laser_link" /> <!-- Set robot laser frame according to your robot -->
    <param name="mapFrameName" type="string" value="v/mjdtwfb31e2e/map" />
    <param name="laserScanName" type="string" value="/r300311695/scan" />
    <param name="absolute_path" type="string" value="/home/sefa/catkin_ws/src/dock_scan_template_matching/src/" />