kunzel / object_search

Lightweight inference tools for hypothesizing object positions and planning viewpoints in 3D space

object_search

Lightweight inference tools for hypothesizing object positions and planning viewpoints in 3D space. The object_search repository is a catkinized ROS package.

Prerequisites

• ROS
• Octomap (1.6)

Please refer to the OctoMap mailing list for installation instructions: 'How to use the latest version of octomap in ROS groovy?'

Getting Started (in simulation)

1. Compile your catkin workspace including the object_search package:

    $ catkin_make
   

2. Open a terminal and run:

   Start the core:

    $ roscore
   

   Launch the simulator: (Please note: the object does not include any objects by default)

    $ roslaunch --wait strands_morse bham_cs_morse.launch env:=cs_lg_obj_search
   

   Add an object in the cs_lg_obj_search.py file as follows:

    cup = PassiveObject('strands_sim/robots/strands_objects.blend','cup')
    cup.properties(Object = True, Type = 'Cup')
    cup.translate(0,0,1.0)
   

   Launch the 2D robot navigation

    $ roslaunch --wait strands_morse bham_cs_nav2d.launch
   

   Launch the octomap server and build a map. Or use a pre-build map:

    $ rosrun octomap_server octomap_server_node `rospack find strands_morse`/bham/maps/cs_lg_obj_search_octomap.bt
   

   Launch the object search utils

    $ roslauch object_search object_search.launch
   

   Run the action server:

    $ rosrun object_search object_search_server.py
   

   Start the search using an action client where <object> is the name of the object the robot is looking for, e.g. Cup:

    $ rosrun object_search object_search_client.py <object>
   

   Start RVIZ and subscribe to a PoseArray named nav_goals and a MarkerArray named /supporting_planes_poses to visualize the viewplanning result.