Create Land Obstacle

[patrick-5546]

Resolves #238

• Created the GeneralPolygon obstacle in the visualizer to represent land
• Created the land_mass_creator.py script to generate land mass input data (see discussion below)

Additional Dependencies

• Shapely

  pip install shapely
  

Usage

The rosparam land_mass_file is a path (relative to this repo's root directory) to a land mass file containing its bounding latlons.

roslaunch local_pathfinding all.launch land_mass_file:=land_masses/land_mass_2.csv

[tylerlum]

Started debugging the issue. Nothing intelligent, just tons of prints with help of vim

Stops at

    state = sailbot.getCurrentState()

Further analysis of prints leads to finding the loop

Essentially, goal waypoint is always invalid. Must be something wrong with the new polygon

[tylerlum]

Played around with it a bit more and got it working! Will let you enjoy the fun of getting it working yourself ;)

[tylerlum]

Great work, very clean code additions

[tylerlum]

Because the above is a potentially infinite while loop, we should add rospy logs.

[patrick-5546]

land_mass_creator.py

Additional Dependencies

• netCDF4

  sudo apt-get update
  sudo apt-get install python-netcdf4
  

Usage

❯ ./land_masses/land_mass_creator.py -h

usage: land_mass_creator.py [-h] [-d THRESHOLD_DEPTH] south north west east

Returns locations of shallow water or land, given coordinates

positional arguments:
  south                 bottom latitude bound
  north                 top latitude bound
  west                  left longitude bound
  east                  right longitude bound

optional arguments:
  -h, --help            show this help message and exit
  -d THRESHOLD_DEPTH, --threshold_depth THRESHOLD_DEPTH
                        maximum depth below sea level of the returned location
                        (in metres, exclusive, default 5)

❯ ./land_masses/land_mass_creator.py 49.115757 49.303264 -123.394709 -123.159747

netCDF dimensions: 48.0020833333 49.4979166667 -124.997916667 -123.002083333
nc index bounds: lat=[267,312) lon=[384,441)
Saved to land_masses/land_depth_5m_10_05_2021-00_02_47.(csv|kml)
Took 0.420228004456 seconds

Outputs

Max depth below sea level of land mass: 5m

• Would this change substantially with the tide?

KML	CSV -> Visualizer

[patrick-5546]

isValid() Performance

Methodology

import timeit
...
class GeneralPolygon(ObstacleInterface):
...
    def isValid(self, xy):
        start = timeit.default_timer()
        is_valid = not self.shapely_polygon.contains(ShapelyPoint(xy[0], xy[1]))
        stop = timeit.default_timer()
        rospy.logfatal('isValid() took {} seconds'.format(stop - start))
        return is_valid

Results

Note: these are upper bounds, average runtime is much faster

• WedgeObstacle: < 5e-05 seconds
• HybridCircleObstacle: < 8e-05 seconds
• GeneralPolygon:
  • land_mass_1.csv (4 points): < 5e-04 seconds
  • land_mass_2.csv (40 points): < 5e-04 seconds
  • vancouver_depth_5m_10_11_2021-10_28_14.csv (84 points): < 7e-04 seconds

Conclusion

The runtime of isValid() in GeneralPolygon (which uses shapely.geometry.polygon.contains()) is an order of magnitude slower than WedgeObstacle. However, it is negligibly impacted by the complexity of a polygon that we would use for a land mass.

[tylerlum]

Impressive results and testing. Great work. Let's get this merged in :)