Skip to content

Using clipper and poly2tri together for robust triangulation

License

MIT, Unknown licenses found

Licenses found

MIT
LICENSE
Unknown
LICENSE.poly2tri
Notifications You must be signed in to change notification settings

acres-com-au/clip2tri

 
 

Repository files navigation

clip2tri

Using clipper and poly2tri together for robust triangulation

About

This code originates from the Bitfighter project (http://bitfighter.org).

We needed a way to generate fast and robust robot navigation mesh zones. Triangulation was a part of this and we had originally used the excellent Triangle library (https://www.cs.cmu.edu/~quake/triangle.html). However, we hope to one day get into the Debian repository and so we needed a more-liberally licensed way to do this. We found the fast poly2tri library in hopes that it would do what we need since no other open source library we found performed as well and had as open a license.

We quickly found that poly2tri has some robustness issues which are inherent to the sweep-line algorithm it uses. Sanitizing our input using the amazing clipper library (http://www.angusj.com/delphi/clipper.php) solves most of the issues.

This code ties it all together.

Details

Bitfighter generates bot nav-mesh zones using the following process (more or less) using floating point data as the input:

  1. Aggregate all polygons and holes
  2. Feed data to clipper and have it return a PolyTree structure
  3. Perform some specific sanitization
  4. Feed data into poly2tri
  5. Feed data into a custom 2D version of recastnavigation (https://github.com/memononen/recastnavigation) to generate the bot zones

The code here handles at least steps 2-4, and has been altered to use more standard data types than those used in the Bitfighter code base.

About

Using clipper and poly2tri together for robust triangulation

Resources

License

MIT, Unknown licenses found

Licenses found

MIT
LICENSE
Unknown
LICENSE.poly2tri

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • C++ 94.6%
  • CMake 3.5%
  • C 1.9%