Hello Cristian and Hello to everyone
I have been blabbering about simulators and decided to write a little one that is functional enough for the community to evaluate and see if useful.
You need to install free cdf player:
https://www.wolfram.com/cdf-player/ (only on Windows and Mac)
You might want to see a paper to get some ideas how the code was organized, p. 2 suffices:
This is a simulator which allows you to enter rotation values for the joints and test to make sure the tips can hold a ball.
The dynamics of the movements are correctly modeled as these papers indicate. Assume the ball is squishy or like a air balloon and the tips are sticky like your finger how they stick to the rubber of a balloon. All kinds of other GRASP could be modeled.
However 20 times a second the position of the ball is computed numerically to assure that it matches the output of the rotary joints.
This is not the way eventually we do the simulators, in this case I cheated and actually solved for the position of the ball using advanced numerical algorithms.
In the final product we will not solve, rather have ways to estimate the solutions with simple matrix multiplications or few such operations without using solvers. Or we train the hand to learn how to grasp… which requires this simulator to generate the training data…
When there are no solutions the ball disappears, you need to back track. This is equivalent of ball falling off your hand or balloon popping since you pinched it hard.
For now this is to show you what I meant by simulators.
Cristian for example could use these sorts of simulators to test his motors and actuators to drive the hand from a CDF file like this. I suggest we build huge numbers of these simulations and provide training and proper documentation for everyone to use.
Use this to learn how to manipulate 3D structures on CDF screen: