Finger Grasp modelling


Hello Cristian and hello to everyone

Some thesis from the Italian geniuses, go to 20 something pages:

Simple 3-joint model:

Springer matrix model, go to (2.28) equations on p25:

For your finger we need a model like these. They look horrific, but they are much easier to understand and code if proper training provided. I can for now model them given a good understanding of your finger.

In 2D/3D you understand how to model motion, any high school student these days understands this:

r: t->(x(t), y(t), z(t))

Then you define equations using r and taking derivatives.

In the case of your finger, r is a matrix space map

r:t->(matrix1(t), matrix2(t), …)

Somehow the matrix() define the orientation of the joints. You solve some differential equations and you are able control the finger (simpler said than done).

I use Mathematica to solve the matrix differential equations, numerically at that, so we can simulate the joint model.

People in the labs make big deal of it for job security, but it is much more manageable if people are honest.



This is a simple grasp model for Cristian to experiment with (and we code a simulator for his robot hand):


  1. “Two matrices are of the utmost importance in grasp
    analysis: the Grasp Matrix G and the Hand Jacobian J” , see p 3.

  2. Ignore all the crap, look at 28.12 and 28.13. Given the design of your hand and servos we can compute the G and J. q-dot is the joint velocities you need to compute on the fly. v is the twist you like to achieve. You find the reverses of G and J and solve for the v and q-dot and feed the information to the servos which then rotate the joints to get the target twist.

  3. Go to p 17 and 18 those computations will be simulated and later on the solutions will be in your micro-controller software to feed the servos.

I will study these and build a hardwired simulator for the two finger grasp of a sphere and see if I could generate the algorithms for Cristians fingers to grasp a sphere in actual setting.


I am studying the 2 finger sphere grasp. I noticed that Cristian’s hand is plastic, he needs to add finger tips with static coefficients over .75:

The above rubber fintertips should do the job



cool! we thought a hybrid silicone-plastic fingers, but that rubber finger grips are easier to put on.


You need to make a prototype hand with 2 fingers, thumb and index to be able to start the software controller.

What I am planning to do is to to simulate the two finger system + rotary wrist.

I called Wolfram Research today, they will help out with the serial/parallel IO interface to your hand from their software.

These are our best ideas at this time to start something here, but you are NOT obliged to accept or cooperate so no pressure.

Nonetheless I will code some simulators.