Skip to content

Latest commit

 

History

History
163 lines (129 loc) · 4.5 KB

README.md

File metadata and controls

163 lines (129 loc) · 4.5 KB
Raw

Search and Rescue Destroy

CS81: Robotics Design and Programming, Fall 2021 at Dartmouth College

Problem:

Map an area and efficiently patrol it, while identifying and pursuing moving objects.

Modules:

Each module may be separately triggered by running the file with the same name as the outer module.

Graph Server (voronoi)

Generates a topological map of the environment from an occupancy grid. See the module-specific README for details.

Patroller

Transforms the topological map into a partol route and dispatched the robot on it.

Camera

Looks for intruders to chase.

Chaser

Takes the robot out of patrol mode and uses a PD controller to chase after intruders.

Restorer

Returns the robot to the patrol route using breadth first search based local planner.

Localization

Determines the robot position on the map, initialtes the state-machine.

Utilities

  • Map publishing
  • Translating JSON-packed topological maps
  • Implementations of BFS, Dijkstra's algorithm

Environment Setup (Gazebo):

If you have a working setup with Gazebo, you can do the following steps to view image data and similar:

Pip Setup:

pip install scipy
pip install numpy
pip install scikit-image

Catkin Setup (Assumes working catkin_ws):

cd ~/catkin_ws/src
git clone https://github.com/husarion/rosbot_description
cd ~/catkin_ws
rosdep install --from-paths src --ignore-src -r -y
catkin_make
source ~/catkin_ws/devel/setup.sh
roslaunch rosbot_description rosbot_rviz_gmapping.launch

# that last command will show some red text. That's OK -- it loads the right world

From here, you have a functional rosbot in a world. If you do

sudo apt-get install ros-melodic-rqt ros-melodic-rqt-common-plugins
source /opt/ros/melodic/setup.bash

rqt
Plugins->Visualization->Image View
then select /camera/rgb/image_raw from the dropdown

you will be able to see image data.

Now run (in separate terminals: you will need 7 for just the nodes. All commands should be run from the root-level directory of this repository) Publish the map

./src/utilities/read_and_publish_map.py

Set up the graph server

./src/voronoi/graph_server.py

Run the LIDAR module

./src/detector/lidar_detect.py

Run the camera module

cd ./src/detector
./camera_detect.py

Run the chaser module

cd ./src/chaser
./chaser.py

Run the patroller module

cd src/patroller
./patroller.py

Run the restorer module

cd src/restorer
./restorer.py

Set up the localizer

./src/localizer/localizer.py

Modifying the World:

Once you've inserted objects into the world you launched with roslaunch rosbot_description rosbot_rviz_gmapping.launch, you can hit File->Save As. You may find that the Gazebo window becomes greyed out and unresponsive. This is because the file saving window actually shows up behind the gazebo window.

You can access the file saving window by minimizing and then maximizing the gazebo window. However, in order to reload the file saving UI, you will need to repeat the minimization+maximization step.

I recommend you type out turtlebot_playground.world and hit enter. This will save the file into your home directory. You should then copy it over to catkin_ws/src/rosbot_description/src/rosbot_gazebo/world. It will override an existing world. When you re-launch the simulation, it will retain your changes.

Building the Map

cd ~/catkin_ws/src
git clone https://github.com/marinaKollmitz/gazebo_ros_2Dmap_plugin
cd ~/catkin_ws
rosdep install --from-paths src --ignore-src -r -y
catkin_make

Add this within the tags of the world file. It will cause the robot to create a map from scan data and publish on the /map topic.

<plugin name='gazebo_occupancy_map' filename='libgazebo_2Dmap_plugin.so'>
    <map_resolution>0.1</map_resolution> <!-- in meters, optional, default $
    <map_height>0.3</map_height>         <!-- in meters, optional, default $
    <map_size_x>10</map_size_x>          <!-- in meters, optional, default $
    <map_size_y>10</map_size_y>          <!-- in meters, optional, default $
    <init_robot_x>0</init_robot_x>          <!-- x coordinate in meters, op$
    <init_robot_y>0</init_robot_y>          <!-- y coordinate in meters, op$
</plugin>

Comparing with Random Walk and A*

Our method traverses the graph in 5min, 10sec

You can run a random walk with

./src/test_drivers/random_motions.py

The random walk took 15 mins, 30 seconds, and failed to explore the center of the map and the room at the bottom.