Torque Measurement for Prosthetic Finger Dynamic Modeling using Neural Network

Hi everyone!

I’ve developed a 15 DOF Prosthetic Arm using Arduino uno and Lynxmotion SSC-32U servo motor controller. Right now I’m trying to model a single finger of the prosthetic arm. When I say modeling, I mean system identification. For system identification I’m using MATLAB’s Neural Network Predictive Controller.

So, firstly I will give a brief about the design overview of the prosthetic finger. Each finger has two joints, that are driven using servo motors placed outside the finger, in the forearm area of the arm linked using cables. Basically each finger is a two link externally driven arm mechanism. Servo motors are used for the actuation and are placed on the foream area. Cables run from the motor to the corresponding joints of the finger in a closed loop. Also, there is a spring loaded slider mechanism on which the motor is housed, whereby each motor can move forward or backwards in the slider to make the cables taut always.

So for moving a single finger 2 data, ie desired angles of the two motors are sent serially to the motor controller. It was understood that the finger was a two input, three output system, where the outputs were the link positions and the torque. Now the torque, at the finger tip need only be considered.

For the modeling, the input and output could be measured. The inputs are given by the user. The output angles of the links we’re measured using two MPU6050 IMU and noting the pitch angles.

For the dynamic modelling of the system I need the torques for a continuous motion.

Guys please guide as to what should be done :smiley-confuse :frowning:

Friction in your drive mechanism may be a large proportion of the forces in something this small.

Are asking how to calculate torques from forces in the control strings?

"system identification" - what do you mean? perhaps "characterization"?