lucia_turtlebot

This repo contains launch + config files to bring up multiple turtlebots including navigation in the Lucia arena, both in simulation and (later) on the real robots.

Installation

Install ROS prerequisites:

sudo sh -c 'echo "deb http://packages.ros.org/ros/ubuntu trusty main" > /etc/apt/sources.list.d/ros-latest.list'
wget http://packages.ros.org/ros.key -O - | sudo apt-key add -
sudo apt-get update -qq
sudo apt-get install -qq -y python-rosdep python-wstool ros-indigo-catkin git build-essential cmake
sudo rosdep init
rosdep update

Install the Las Vegas Reconstruction toolkit:

sudo apt-get install -qq -y libfreenect-dev libopencv-dev libflann-dev libeigen3-dev libvtk5-dev libvtk5.8-qt4 python-vtk libvtk-java libboost-all-dev freeglut3-dev libxmu-dev libusb-1.0.0-dev
mkdir -p ~/lucia_software
cd ~/lucia_software
git clone https://github.com/lasvegasrc/Las-Vegas-Reconstruction.git
mkdir -p Las-Vegas-Reconstruction/build
cd Las-Vegas-Reconstruction/build
cmake -DCMAKE_BUILD_TYPE=Release ..
make
sudo make install

Install the 3D Toolkit 3DTK:

sudo apt-get install -qq -y libzip-dev libann-dev libsuitesparse-dev libnewmat10-dev subversion
cd ~/lucia_software
svn checkout svn://svn.code.sf.net/p/slam6d/code/trunk slam6d-code
mkdir -p slam6d-code/build
cd slam6d-code/build
rm ../3rdparty/CMakeModules/FindCUDA.cmake
cmake -DWITH_FBR=ON ..
cmake -DWITH_FBR=ON ..
make

Be sure to make the cmake -DWITH_FBR ... call twice to get the software compiled.

Create a Catkin workspace, clone and build our ROS stacks:

mkdir -p ~/catkin_ws/src
cd ~/catkin_ws/src
wstool init -j4 . https://raw.githubusercontent.com/uos/uos_rosinstalls/master/lucia2016-indigo.rosinstall

# Use rosdep to install all dependencies (including ROS itself)
sudo apt-get install -qq -y linux-image-extra-$(uname -r)  # workaround for ros-indigo-realsense-camera
rosdep install --from-paths ./ -i -y --rosdistro indigo
sudo apt-get install -qq -y ros-indigo-rviz ros-indigo-rosbash

source /opt/ros/indigo/setup.bash
catkin_init_workspace
cd ~/catkin_ws

catkin_make -DCMAKE_BUILD_TYPE=Release

If you have rosjava installed, you may have to repeat the catkin_make command several times until it succeeds. The generated setup.bash file has to be sourced in each terminal that uses this catkin workspace. It's best to add it to the .bashrc so that it is sourced automatically whenever a new terminal is opened:

echo "source ~/catkin_ws/devel/setup.bash" >> .bashrc

Now close all running terminals and open a new one.

Updating

To update to the latest version of all packages, do this:

cd ~/catkin_ws/src
wstool merge https://raw.githubusercontent.com/uos/uos_rosinstalls/master/lucia2016-indigo.rosinstall
wstool up
rosdep install --from-paths ./ -i -y --rosdistro indigo
cd ~/catkin_ws
source ~/catkin_ws/devel/setup.bash
genjava_message_artifacts
catkin_make -DCMAKE_BUILD_TYPE=Release

If you have rosjava installed, you may have to repeat the catkin_make command several times until it succeeds.

Setting up your Eclipse workspace for the MetaCSP tutorial (Wed)

cd ~/catkin_ws/src/lucia16/metacsp_tutorial
./gradlew eclipse

Go to Eclipse:

• make a new Java Project
• unclick "Use default location"
• Browse to ~/catkin_ws/src/lucia16/metacsp_tutorial/multi_robot_active_perception/
• Click Finish

Instructions for the OWL-DL tutorial (Tue)

roscd lvr_plane_classifier/
git pull
roscore

# start rviz in a new terminal:
rosrun rviz rviz -d $(rospack find lvr_plane_classifier)/rviz/semantic_furniture_recognition.rviz

# start semantic_furniture_classifier_node in a new terminal:
roscd semantic_furniture_classifier/semantic_furniture_classifier_project/
build/install/semantic_furniture_classifier_project/bin/semantic_furniture_classifier_project com.github.semantic_furniture_classifier.semantic_furniture_classifier_project.SemanticFurnitureClassifierNode

# start lvr_plane_classifier_node in a new terminal:
roscd lvr_plane_classifier/
wget http://kos.informatik.uos.de/lucia_data/tableandchairs.zip
unzip tableandchairs.zip
mkdir -p /tmp/tableandchairs
rosrun lvr_plane_classifier lvr_plane_classifier_node tableandchairs/tableandchairs1.ply

# instead of restarting the lvr_plane_classifier_node over and over, you can later run this to save time:
roscd lvr_plane_classifier/
rosbag play --clock -d 1 tableandchairs/tableandchairs1.bag

• Download Protégé 5.1.0, unzip, start run.sh.
• Enable Window -> Tabs -> SWRL Tab.
• Open ~/catkin_ws/src/semantic_object_maps/semantic_furniture_classifier/semantic_furniture_classifier_project/owl/furniture.owl.
• Look around in the "Entities" tab. You don't need to change anything here.
• Open the SWRL tab. Your goal is to add rules to classify a chair. For this, you need to add two rules: one that classifies certain HorizontalPlanes as ChairSeatPlanes, and one that classifies certain VerticalPlanes as ChairBackrestPlanes. If you want to take measurements, you can either use the "Measure" tool in RViz, or simply take a ruler to the actual chairs.
• Save the ontology file.
• Restart semantic_furniture_classifier_node and rosbag, and look at the RViz markers. When two ChairSeatPlane labels appear, you have succeeded! :)

Instructions for the FLAP4CAOS tutorial (Tue)

Your task: Develop your own search strategy for active perception (see slides). In short:

• Follow the instructions below to start everything in Gazebo with only one turtlebot.

• Watch the results in RViz.

• Cancel everything again.

• Edit ~/catkin_ws/src/uos_active_perception/race_object_search/src/search_planner_tutorial.h to implement a brilliant new search strategy.

  • All that is required is that makePlanRecursive() returns a sequence of actions.
  • The first action will be executed, then the planner will be called again, so returning only the next action is also valid (if you want to implement a more reactive approach instead of planning to the end).
  • To give your planner more time, you can disable the timeout in the code or in ~/catkin_ws/src/lucia_turtlebot/lucia_launch/launch/includes/object_search_manager.launch.

• Then:

cd ~/catkin_ws
catkin_make -DCMAKE_BUILD_TYPE=RelWithDebugInfo --pkg race_object_search

• Rinse, repeat.

Running in Gazebo

roslaunch lucia_launch lucia_gazebo.launch

Now, you can use the "2D Nav Goal" tool in RViz to send navigation goals to the first turtlebot. If the robot moves to the goal position, you have everything set up correctly. You can command the other two turtlebots by opening "Panels" -> "Tool Properties" and changing the "2D Nav Goal" topic.

You can disable specific turtlebots like this (for example, to speed up the Gazebo simulation):

roslaunch lucia_launch lucia_gazebo.launch turtlebot2:=false turtlebot3:=false

To start FLAP4CAOS active perception, run this:

ROS_NAMESPACE=turtlebot1 rosrun race_object_search object_search_manager_test p lucia_area_sw 0.4 lucia_area_nw 0.4 lucia_area_e 0.4 min_p_succ 0.05

Trying out your code with the Turtlebots

Relevant addresses and other info

• Basestation: 10.0.0.200

• Turtlebot laptops:

  • perception1 (10.0.0.211 iran/admin123)
  • perception2 (10.0.0.212 robot/robot)
  • perception3 (10.0.0.213 robot/robot)

Instructions

• Start the following on basestation

roscore
roslaunch lucia_launch rviz.launch

• For the FLAP4CAOS tutorial (Tue): make sure your code is pushed to your group's Git repo, checked out and compiled on perception1-3

• Start the following on perception1-3

roslaunch lucia_launch lucia_robot.launch robot:=turtlebot1
roslaunch lucia_launch lucia_robot.launch robot:=turtlebot2
roslaunch lucia_launch lucia_robot.launch robot:=turtlebot3

• Unplug the turtlebots you are using, place them in the arena and provide pose estimate for each via Rviz
• Send a 2D nav goal for each robot
• Start the following on perception1-3

roslaunch lucia_launch perception_laptop.launch turtlebot1:=true
roslaunch lucia_launch perception_laptop.launch turtlebot2:=true
roslaunch lucia_launch perception_laptop.launch turtlebot3:=true

• For the FLAP4CAOS tutorial (Tue): also start the following on perception1-3

ROS_NAMESPACE=turtlebot1 rosrun race_object_search object_search_manager_test p lucia_area_sw 0.4 lucia_area_nw 0.4 lucia_area_e 0.4 min_p_succ 0.05
ROS_NAMESPACE=turtlebot2 rosrun race_object_search object_search_manager_test p lucia_area_sw 0.4 lucia_area_nw 0.4 lucia_area_e 0.4 min_p_succ 0.05
ROS_NAMESPACE=turtlebot3 rosrun race_object_search object_search_manager_test p lucia_area_sw 0.4 lucia_area_nw 0.4 lucia_area_e 0.4 min_p_succ 0.05

• For the MetaCSP tutorial (Wed): Connect your laptop to LuciaSchool2 network (password on the whiteboard)
• For the MetaCSP tutorial (Wed): add the following to your .bashrc and source it

ROS_MASTER_URI=http://10.0.0.200:11311
ROS_HOSTNAME=<your-IP-address>

Package overview

lucia_launch

This ROS package contains all top-level launch files:

• lucia_gazebo.launch - Brings up everything on 3 Turtlebots in Gazebo.
• lucia_robot.launch + perception_laptop.launch - Same as above for physical robots.

lucia_turtlebot_description

URDF model of the modified Turtlebots used in the Lucia school (i.e., regular Turtlebots with a pole and second Xtion added).

lucia_gazebo_worlds

Gazebo world, model and launch files for the Lucia arena.

multi_turtlebot_navigation

Launch files to bring up navigation (move_base, amcl) on multiple turtlebots.