（1）Basic idea：map segmentation
（2）The concept of Cell is important in Coverage. It represents the minimum unit in the map. For some situations, the developers can determine the shape of Cell for any kinds of agents. Square is the most common one for coverage.
（3）For the cost_map library, grid map is the important format for avoiding the obstacles. In the common map, we will describe the position by (x,y). But in the grid map, we describe the position by the index of a cell. For example, the (x,y) in the common map will be converted into the (x*width+y) in grid map.
（4）2019.07.06 the demo has been tested successfully, but it still shows some problems in this test:
a.There is some long lines in the costmap which has been planned. It is impossible for the agent to run.
b.In this algorithm, a lot of turns were produced, that will cost a lot of energy for our agents.
（5）parameters adjustment
（6）For solving the self-locking problem , I add a new method in the original algorithm. A* algorithm(point-to-point path plan) is adopted for this problem.
（7）Reducing the number of turns. Change the scoring mechanism.
（8）If you want to run this program on your devices. You need to install some dependences.
I show the results of this program as follows:
After adjusting the algorithm, the number of turns has been reduced substantially. But the path on oblique section is not ideal. This phenomenon may be caused by the meshing of our map.
Following is the result of reducing turns:
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
If you find some disadvantages about this program, or you have a better idea to achieve the coverage and improve this program, welcome to discuss!
By the way:
For realizing the path-coverage algorithm, we have to align the frame between map and the real world. In the next, I will describe some special "origins" in the map frame.
In the above map, the bottom-left corner is the origin (0,0) of map frame, the top-left corner is the origin of Mat in CV, the "col" and "row" of the top-left corner are both 0.
In the config files, we need to adjust two key parameters---size of cell, resolution. Size of cell determines the size of meshing for the map. For example, size of cell = 21, it means the size of a meshing cell is 21x21(pixel). The resolution represents the size of a pixel. For example, resolution = 0.5, it means the size of a pixel is 0.5x0.5. We need to notice that "0.5" has the same scale with map frame.