This fork was only tested with ROS Noetic.
A reworked version of the original TMSTC_Star package was implemented as a service that takes a map, initial robot positions, and the cleaning tool width and returns a list of paths. Mostly identical to original results, however, the grid resizing is now done using the grid_map package and uses opencv.
I have also completely removed the turtlebot simulation since this was not needed here and should be decoupled from TMSTC* in a demo package.
It partitions the map with a minimum number of bricks as spanning tree’s branches, greedily connects bricks to form a complete spanning tree and then applies the greedy strategy of MSTC* to find the optimal equilibrium division on the topological loops around the spanning tree, thus averaging the weights of each robot coverage path. In addition, it incorporates turn cost into the weights of the paths.
cd ~/catkin_ws/src
git clone https://github.com/MDP-Group-16/TMSTC-Star.git
cd ../
rosdep install --from-paths src --ignore-src -r -y
catkin_make
source ~/catkin_ws/devel/setup.bash
(maybe you alsoe have to sudo apt-get install libopencv-dev before, not sure)
roslaunch TMSTC_Star test.launch
/allocate_method: "DARP", "MSTP" (broken) or "MSTC"
/MST_shape: "DFS_HORIZONTAL", "DFS_VERTICAL", "KRUSKAL", "ACO_OPT" (broken), "HEURISTIC", "OARP", "RECT_DIV" or "DINIC"
/useROSPlanner: use the ros planner for point to point navigation (currently only false works)
/coverAndReturn: let robots return to its depot or not, "true" or "false"
/free_threshold: at what point is a cell considered occupied (ie will not be cleaned) after interpolation (0-100, 100 is occupied)
If sets "MSTC" and "DINIC" at the same time, the planner will perform TMSTC*. Change main.launch to add more robots. If you want to use anotheor map, remember to modify image path in map/map.yaml as well.
mCPP on various environments:
