Kinematic artificial neural network Javascript library
npm install kinannKinann lets you create an artificial neural network (ANN) that bridges
the gap between a simple, ideal kinematic model and any given implementation
of that kinematic model. Kinann will handle all the error corrections
automatically after proper calibration and training. Kinaan doesn't actually
need to know the precise kinematics of your model--all it does is model
the mismatch between ideal and actual coordinates. As long as your robot
is precise, Kinann will make sure that your robot moves accurately to
application coordinates:
IdealKinematics + Kinann = CalibratedRobot
Kinann kinematic error regression ANNs can be linear or even polynomial.
Linear Kinann networks are often sufficient for Cartesian kinematics. However,
you will need polynomial Kinann networks to deal with non-linear kinematics.
For example, rotary delta kinematic errors often manifest as "bowl-shaped Z-plane errors".
Kinann is designed for building kinematic neural networks for Javascript robot applications.
Although there are many neural network frameworks
(e.g., synaptic.js,
Tensorflow, etc.) that can be used to solve
the kinematic error challenge, Kinann is optimized for kinematic modeling.
Kinann should not be used to recognize cat pictures on YouTube.
npm to install kinann.npm install kinann
* Kinann wiki...
* mathjs many thanks to MathJS for expression parsing and derivatives!