kinann and FPD

[Karl]

Out of curiosity, I took a look at FPD kinematic calibration based on Kinann, the FirePick neural network framework.

The kinematic equations for rotary deltas are a bit big and I wanted to see how Kinann performed with a tougher

but very real kinematic equation system. Unfortunately, I hit a brick wall with mathjs derivative() performance, which

didn't compute after several minutes of waiting. I am currently working with Jos to see how we can extend mathjs

to address Kinann's needs. We do have a solution in hand, but our next challenge is mathjs integration of this new

functionality.

https://github.com/josdejong/mathjs/issues/811

[Karl]

Kinann 0.3 includes the Equations class, which addresses the mathjs performance issue for Kinann. Jos is looking into mathjs integration of Equations, but Kinann is moving forward independently. With Equations, I hope that Kinann can apply neural networks to the FPD calibration problem.

https://github.com/firepick/kinann/releases

However, since Felix expressed an interest in working together on Cartesian machines, I'm putting FPD on the backburner again. The neural network approach requires some redesign of firenodejs, and I'll be looking Kinann integration with firenodejs