/peg_in_hole_project

Primary LanguageCMake

peg_in_hole_project

filter.launch

launches the following:

  • Force torque sensor node, published wrench (ros package net-ft-ros)
  • Python FT classifier, published feature labels (ros package ft_sensor_python)
  • Virtual peg_sensor model, used to estimate expected measurements (not necessary when using FT) (ros package peg_sensor)
  • Particle filter (ros package peg_filter)

peg_filter

The peg filter node does two things; 1) update the particle fitler and 2) compresses the particle filter to a lower dimensional representation. It published the particle filter weights which are shown in Rviz and the compressed feautres as well.

publish

topics: [name,type]

  • [mode_feature, std_msgs::Float64MultiArray>]
  • [bel_x,visualization_msgs::Marker]
  • [pfilter,sensor_msgs::PointCloud]

server.launch

Initialised and launches the server actions. One action, plug_search, is added to the action server.

Action plug_search (ros package Peg_hole_policy)

peg_kuka_cmd.launch

  1. call plug_search
  • startes the plug_search action in the action server
  1. call gmm
  • starts the GMM policy via a call to a service in the plug_search action.

publish

  • actions (velocity) to the robot controller

subscribe

topics: [name, type]

  • [ft_classifier,std_msgs::Float32MultiArray]

Sensing

  • virtual sensor topic: /virtual_classifier : [std_msgs/Float64MultiArray], distance to features
  • ft sensor topic: /ft_classifier : [std_msgs/Float64MultiArray]

Useful things

  • rosrun rqt_reconfigure rqt_reconfigure

Demos