Coverage path not working when the footprint is very small

[Adrianrr94]

Greetings,
I have been using your ROS service for a project where the drone need to do a coverage path but the footprint will be really small, like 5-6 m (vertical and horizontal).
So, to debug the code, I tried your specify_rect.py to see if the problem was in my code or in your ROS service.

To do that, I changed the parameters on the specify_rect.py's interface in order to have a small footprint, obtaining as a result that there is a zone without plan as it can be seen in the screenshot.

That's exactly the problem I am having in almost all my planification, I believed this is due to the fact that the footprint is small and the area to cover very big. I would like to know how to fix that.

Thanks.

[uenota]

Thank you for reporting the problem. I will check the code.

However, I have not worked on this project for a long time, so it may take long time to find out the causes.

[UTWDrones]

Hello @uenota
Debuging your service ("torres_etal_2016.cpp"), I found out that when the footprint if small and the given polygon to coverage is a triangle the method computeConvexCoverage returns False no matter where the start points is. So the triangle is decomposed and so that, an error occurs.

`// isOptimal is true if computed path is optimal
bool isOptimal = computeConvexCoverage(polygon, footprintWidth.data, horizontalOverwrap.data, candidatePath);
if (isOptimal == true)
{
// fill "subpolygon" field of response so that polygon is visualized
res.subpolygons = generatePolygonVector(polygon);
// set optimal alternative as optimal path
// see torres et al. 2016 for Optimal Alternative
res.path = identifyOptimalAlternative(polygon, candidatePath, start);
}
else
{
std::vector subPolygons = decomposePolygon(polygon);

// sum of length of all coverage path
double pathLengthSum = 0;
// compute length of coverage path of each subpolygon
for (const auto& polygon : subPolygons)
{
  PointVector partialPath;
  computeConvexCoverage(polygon, footprintWidth.data, horizontalOverwrap.data, partialPath);
  pathLengthSum += calculatePathLength(partialPath);
}
// existsSecondOptimalPath is true if there is at least one coverage that has no intersection with polygon
// second optimal path is the path that has second shortest sweep direction without any intersection with polygon
PointVector secondOptimalPath;
bool existsSecondOptimalPath =
    findSecondOptimalPath(polygon, footprintWidth.data, horizontalOverwrap.data, candidatePath);
if (existsSecondOptimalPath == true)
{
  // compute optimal alternative for second optimal path
  secondOptimalPath = identifyOptimalAlternative(polygon, candidatePath, start);

  // if the length of second optimal path is shorter than the sum of coverage path of subpolygons,
  // set second optimal path as the path
  if (pathLengthSum > calculatePathLength(secondOptimalPath))
  {
    // fill "subpolygon" field of response so that polygon is visualized
    res.subpolygons = generatePolygonVector(polygon);

    res.path = secondOptimalPath;
    return true;
  }
}
else if (subPolygons.size() < 2)
{
  // if number of subpolygon is smaller than 2,
  // it means no valid path can be computed
  ROS_ERROR("Unable to generate path.");
  return true;
}

`
I am not sure this is the part of the code that is making the algorithm to fail, but it maybe the cause.

I look forward to hearing from you.
Thanks.

[uenota]

@UTWDrones
Thank you for feedback!
I fixed the issue about computeConvexCoverage.

However the issue @Adrianrr94 reported is not fixed yet.
I will continue to investigate the problem.

[uenota]

@Adrianrr94
The issue you reported should be fixed in e526ab8.
Could you check if it fixes your issue?

I'm going to close this issue. Please reopen if the problem persists.