This is the knowledge implementation of the SUTURO1819 project.
Use wstool in a seperate workspace to install the dependencies from dependencies.rosinstall, including iai_common_messages and knowrob.
Run the Prolog interpreter with rosrun rosprolog rosprolog object_state.
To run the beliefstate, run
roslaunch object_state object_state.launch
For the URDF of the environment, choose the following:
roslaunch iai_hsr_robocup hsr_robocup_with_state_publisher.launch
or if you want to run the URDF and map together, do
roslaunch iai_hsr_robocup hsr_maps.launch
which launches the map and snap_map from the hsr_navigation, and the URDF state publisher of the environment.
Create one catkin workspace for the dependencies and load the wstool file into the src directory
# to setup basic catkin workspace
source /opt/ros/kinetic/setup.bash
rosdep update
mkdir -p knowledge_deps/src
cd knowledge_deps
catkin build
# would recommend build over catkin_make. install catkin_tools via apt if needed
# now load the dependencies.rosinstall into your wstool config
cd src
wstool init
wstool merge https://raw.githubusercontent.com/Suturo1819/hsr_knowledge/master/dependencies.rosinstall
wstool update
rosdep install --ignore-src --from-paths .
cd ..
catkin build
The build will take some time, and if it fails, use source devel/setup.bash to get the current build. The run catkin build again.
Now, in another workspace you can load this repository.
source /path/to/knowledge_deps/devel/setup.bash
cd /path/to/all/the/workspaces # it's /home/suturo/robocup in our case
mkdir -p knowledge_ws/src
cd knowledge_ws
catkin build
#clone this repo
cd src
git clone https://github.com/Suturo1819/hsr_knowledge.git
cd ..
catkin build
source devel/setup.bash
You should now have two workspaces, one named knowledge_deps and the other one knowledge_ws. You will probably never need to build the _deps workspace again, but keep updating hsr_knowledge in the _ws workspace occasionally and use catkin build to get the new functionality into your environment.
Don't forget to put the source commands with absolute path into your .bashrc if your want to source the workspace in each new terminal. Remeber also, that the last source command in your .bashrc always overwrites all the previous ones. Look into workspace overlaying for further explanation.
Launch the robot, since the object state ublisher needs a map frame. The hsr can be in standby though.
roslaunch object_state object_state.launch launches the beliefstate of the hsr objects, the object_state_publisher and the perception subscriber. The subscriber listens to the topic /hsr_perception/results, decodes the message and asserts it into the beliefstate.
Through json_prolog queries you can obtain the data:
object_at_table(Instance). provides all objects at the approximate position of the table (-1, 1, 0.8).
object_of_type(knowrob:'Cup', Instance). gives all objects of type Cup. Replace Cup with your favourite class.
A more adventurous predicate can be useful to find objects at different locations:
object_at(knowrob:'Cup', ['map', _, [-1,1,0.8],[0,0,0,1]], 0.4, Instance). This will give you all the instances of type Cup within a 0.4 meter distance from (-1, 1, 0.8). Make the first argument a wildcard _ if you don't care about the type.
Use the ROS-Protege version from Daniel, to load ontologies, that reference ros-packages. https://github.com/Suturo1819/ros-protege.git
rosrun object_state perception_action_client.py hsr_perception_<table,shelf>