Thanks a lot! It would be cool to have some of it get used! I've been trying to get all of it online. Right now there is not really anything coherent ready to put up yet - a smattering of Jupyter demos using/explaining aspects of the code. this keeps getting triaged for another day, and yet I keep seeing bezier and B-spline stuff get popular on here.
It's a huge swath of messy python with the first stuff dating back to 2012. I used Piegl and Tiller's book C++ converted to python for the underlying B-spline stuff. (needed to implement anyway in order to compute constraints and objectives for custom auto-diff solver...)
Then convert the spline control points into an overloaded automatic differentiation type for use in construction of the basic curve solver. That we at least have documented publicly in the journal of CAGD here: https://www.sciencedirect.com/science/article/abs/pii/S01678...
The stuff above made it easy to be really flexible about what constraints are used in any given curve. (Or surface, but I did not go there for direct optimization.)
Next comes... all the interval arithmetic and relational constraint logic to pare down the design space to only the feasible domain on the front end. It's going to take a while to clean up the code and tell the story better.
The trick for kayaks is that my professor had me focus on industrial applications. -Designing offshore supply vessels, that sort of thing.
I could work with you to try and re-tool a bit to see what we could get. I can't make promises though! I am swamped with a new baby and various other job related things at the moment.
Nevertheless it should be pretty easy to get started, since a kayak is a sort of generalized canoe, which is the first hull form anybody generates. The outstanding issue is that I've not yet worked on developable constraints!
(Actually I've been tooling up to bring some discrete differential geometry to bare on this) The issue here is that nonlinear B-spline surface optimization direct is expensive. See, e.g. https://www.cs.cmu.edu/~kmcrane/
Developable triangle meshes and other such papers. Cool stuff. Of course the shape representation is different.
There are the old directrix techniques as well.
It's a huge swath of messy python with the first stuff dating back to 2012. I used Piegl and Tiller's book C++ converted to python for the underlying B-spline stuff. (needed to implement anyway in order to compute constraints and objectives for custom auto-diff solver...) Then convert the spline control points into an overloaded automatic differentiation type for use in construction of the basic curve solver. That we at least have documented publicly in the journal of CAGD here: https://www.sciencedirect.com/science/article/abs/pii/S01678...
The stuff above made it easy to be really flexible about what constraints are used in any given curve. (Or surface, but I did not go there for direct optimization.)
Next comes... all the interval arithmetic and relational constraint logic to pare down the design space to only the feasible domain on the front end. It's going to take a while to clean up the code and tell the story better.
The trick for kayaks is that my professor had me focus on industrial applications. -Designing offshore supply vessels, that sort of thing. I could work with you to try and re-tool a bit to see what we could get. I can't make promises though! I am swamped with a new baby and various other job related things at the moment.
Nevertheless it should be pretty easy to get started, since a kayak is a sort of generalized canoe, which is the first hull form anybody generates. The outstanding issue is that I've not yet worked on developable constraints! (Actually I've been tooling up to bring some discrete differential geometry to bare on this) The issue here is that nonlinear B-spline surface optimization direct is expensive. See, e.g. https://www.cs.cmu.edu/~kmcrane/ Developable triangle meshes and other such papers. Cool stuff. Of course the shape representation is different. There are the old directrix techniques as well.