http://log.bondhan.web.id

I intend to put anything related to technology on that site, So that It Reminds Me (TM) always.

However this blog will be as it is. It has become my notes during my years in Korea, so many happened during those days. Although I am not going to update this blog any longer (for now), I am still reading the coming comments and try to help if possibilities are there.

Thank you for your coming. Cannot wait to see you at my new home

I’d like to thank for everyone who has helped me during my stay in Korea.

The details are on my thesis, please note that the data is in preliminary test.

Reference:

A 6-DOF Gait Rehabilitation Robot with Upper and Lower-Limb Connections that Allows Walking Velocity Updates on Various Terrains. Jungwon Yoon, Bondhan Novandy, Chul-ho Yoon, and Ki-Jong Park. IEEE Transactions on Mechatronics. Focused Sections: Healthcare Mechatronics. 2010. (Accepted)

Design scheme:
First, select a PI sliding surface. Second, from the sliding surface we derive the equivalent control. Third, the control output is achieved by adding the switching control output. The mathematics derivation is straight forward.

Here is an experiment of position control using Proportional sliding mode control of a BLDC motor. Not to be in a formal dress, you can download the presentation file here.

PISMC Setup

Video:

For the first trial, we have the IXXAT CAN-USB device. The question is, how to interface USB device to the RT? Because if I use this device, then it will be RT <-> USB <-> CAN, which means that I need a driver for IXXAT CAN-USB for the RT! I emailed the IXXAT company, the engineer replied that they do not support the LV-RT. Anyway I successfully send command through CANOpen with this device on Windows platform, but it took 1 day for me to realize that I need to terminate the final CAN-bus with 120 Ohm. Fiuhh.. alhamdulillah it was working though

Then we bought the NI-CAN PXI 2 Port from NI with the OPENCan library. I successfully controlled the speed of the BLDC motor with the CANOPen using SDO only, but then I realized that there was a big delay between two SDO commands, which resulted almost 100% error (current speed – desired speed), currently I am going to write a new program by using both SDO and PDO.

I -insyaAllah- will continue this writing and explain briefly about SDO & PDO especially in the case of Maxon EPOS.

Monday, 4/feb/2007 @ 20:52 KST

The definition of CAN and CANOpen can be found by yourself in google . I will explain briefly what I have learned, the CANOpen method to control the Maxon Motor using EPOS, for more detail you can read the EPOS Application Note: CANOpen Basic Information (pdf).

In EPOS there is an important dictionary (the source of everything), called Object Dictionary Access. It contains the System parameter (e.g. PID tuning value, velocity, etc), the DataRecorder(Encoder counter, etc), and so on. To control the motor, these parameters need to be changed. We can change the value by 3 methods:

1. Directly click the field (of the column) and enter the desired value
2. Using the Serial communication RS232 protocol
3. CANOpen protocol

In the case of 1, you must be joking, but if you want to do it manually, no one can blame on you =P. The 2nd method, I have explained above, that it is not possible if we want to utilize the LabView RT, because there is a dll file which does not support the RT so it can not be downloaded onto the RT, unless you know how to reverse engineer this file and re-write your own code .

Now the 3rd method, with CANOpen there are 2 ways of writing the object dictionary access:

#1 is SDO (Service Data Object). It is a confirmed service, means that after you send the command, in between the timeout value you specified, there will be a confirmation packet which contains the value whether you successfully write the object or not.

#2 is PDO (Process Data Object). Opposite to the SDO, it is unconfirmed, thus it is very fast. But as a trade off, there are a “little” work to do, which is the PDO mapping.

At first time, I am using the SDO only, the problem is that, there is a big gap between sending the packet and receiving the confirmation packet, thus it become less real-time and worse, the data is not valid (I am writing a quite time critical program) .

Here’s some picture with SDO only:

If you take a look at the picture above the error looks very big, I guess it is caused by the delay of the SDO packets.

And in the above picture the error is very small with PDO packets.

And here is the video, if you need the code just write me a comment or send an email

This is an example program to initialize the Maxon EPOS through CANOpen. To be noted that the maxon UI Interface is always on when the lv program is running. Download here (rename the file to become .zip)!

Update: Thu 29/09/2011

Our machine configuration:

NI-PXI 7358 Motion Controller, NI-PXI 6259, NI-CAN PXI-8461/2 + CANOpen Library, Maxon BLDC Motor + EPOS 70/10, DES 70/10

The Matlab file will not be shared, I think the simulation is not difficult you can write in S-Function or by calling ODE solver from m-file. LabView source code can be seen (seeing source code or reading it ) in the presentation file. I hope It is useful for the sake of education, sharing knowledge to the world -at least this is what I can do for now , hope I could give something useful to human being before I die-. Btw, if there’s something wrong with the equation please let me know. Oh and..don’t expect too much from the presentation file, I only did this less than 2 weeks, plus I am new to LabView and Matlab S-Function T_T.

After the encoder is installed to the motor I will post the video here, insyaAllah. Hope It will not take too much time for the experiment.

Updated 2008-05-07:

I did the experiment last february 2008, just able to upload it now caused by my laziness . I forgot to add in the presentation file, the inverter input is only positive DAC value from 0 to 10 V. Thus, if the control output is negative you need to make additional circuit to switch the direction (from forward to reverse, vice versa). (Look at my posting about the “Basic Switching Circuit”), also for safety you must add switches and hardwired to the inverter. Here are the setup and the video:

SMC AC Servo Motor experiment

Video:

Update 2008-12-08

The simulation file can be downloaded here (rename to .zip), written on LabVIEW 8.5, all the derivation can be found on the pdf (ppt).

Update 2010-01-15 (OMG, more than a year already :p)

I admit that output response in this video was not that good since it was not tuned (I am lazy to upload the new one). The final result, after tuning the control parameters, the whole result can be seen in here. It is fully programmed using LabVIEW & LabVIEW (RT).

Here’s the code when I did a measurement on the time of digital I/O switching on-off:

I decided to utilize the PWM port. The algorithm is: turning on 100% PWM -> delay X uS -> turn on 0% PWM -> delay 10 ms (minimum), where X is the delay for the position desired, and then the servo could run smoothly.

Here’s the code:

As usual, for documentary (maybe when I am old I could see what I did in my “spare” time -hopefully not wasting my time-), here’s the video:

Two days ago, I tried to control a stepper motor with ULN2803 and LabVIEW. FYI -again- I am new to LabVIEW, so don’t expect too much from this article. I have NI-PCI 6221 and NI-PXI 7358 with me, the later is specifically manufactured for motion controller but still it has the digital I/O. And in this article I am using the motion controller digital I/O to control the stepper, if you only have DAQmx you also still can use it to drive your stepper motor, because to control stepper motor only need digital output which you can generate from your DAQmx.

Here’s the LV source code:

This is the front panel where user can specify the speed, direction, number full step. In this case, I am using 1.8 degree stepper motor, which means for one full rotation I need 360/1.8 = 200 steps, then 200 steps/4 = 50 times iteration.

This is the source code when all condition is true:

This is the source code when all condition is false:

Here’s the video for the trial:

and I tried to test the accurateness of this stepper motor for one full cycle (50 iterations) you can watch here:

This stepper motor is not that accurate, for 2-3 trials it was okay but the next trial it lost its position, I tried several times and it still can not maintained its previous position, even though I used slower speed.

And here is the electronic connection I used for the ULN2803 you can find here.

Oh well, I decided to learn how to generate a simple PWM in PCI-6221 on LabView -as you know I am always new in this, too many functions available, and I am not used to graphical programming-.

I utilized the counter 0 (DevX/ctr0) to turn on a LED (light emitting diode) that it will flicker according to the signal given. For this card the LED is located on PFI 12 (pin number 2).

Here is the front panel, where user can interact with the program:

and here is the back panel for the main flow program:

I still not yet have the LabView Real-Time with me, my boss said it will arrive tomorrow. This program is written with LV 8.2 with the minimum delay is 1 ms. If you analyze the back panel above, when there’s a value change in the control form (the meter and vertical bar) for different duty cycle or frequency, the value will be updated inside the case structure. You can experiment by changing the duty cycle and frequency, and then look at the flicker rate and intensity of the LED.

If you removed the timed loop and changed with while loop, a problem will occur if the value change in the control form is faster that one period of loop cycle, thus LabView Run-Time will generate an error message. The reason is that, the changing is lot faster that one cycle loop which is 1 ms. With this timed loop, unnecessary changing is discarded, only values which are between the right time slot considered as input. If error happened the LED on front panel will be on.

If you don’t understand my explanation, you can download the vi program here (remove the .pdf extension).

I am not going to explain the whole theory, I will just show you the homework and Matlab program that I wrote for this transformation type. In Screw transformation, the reference and final point is arbitrarily chosen, this makes this method different to DH method.

Ok. Here’s the question:

Here’s the answer:

First we find transformation matrix using screw method. The screw axes location of Scorbot manipulator is shown in Table 1, then we define the reference, target position and then we can find the transformation matrix. With this transformation matrix we can derive the inverse kinematics.

a) Reference Position

u₀ = [0, 0, 1]^T, v₀ = [0, -1, 0]^T, w₀ = [1, 0, 0]^T

and p₀ = [a1 + a2 + a3, 0, 0] ^T

b) Target Position

u = [u_x, u_y, u_z]^T, v = [u_x, u_y, u_z]^T, w₀ = [w_x, w_y, w_z]^T

and p = [p_x, p_y, p_z]^T

c) Transformation Matrix

By using Matlab® (See the attached code) the transformation matrices are found (Use the matlab code)

d) Inverse Kinematics

You can derive the inverse kinematics using the direct kinematics matrices calculated using the Matlab program, not that difficult, so I will leave this solution to you.

Matlab(R) 2007a source code:

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Homework chapter 2 Number 11
% Book: Robot Analysis
% Name: Bondhan Novandy
% Student ID: 2007214012
%
% Description:  This file describe the screw tranformation, the input is the
%               screw axis location coordinate Si and Soi (table), the
%               output is the transformation matrix
%
% Dependency :  None
% Note       :  Tested on Matlab(R) 2007a
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

function [output] = screw_transform(val_sx, val_sy, val_sz, val_sox, val_soy, val_soz, symTheta,  joint_type)

global REVOLUTE PRISMATIC
REVOLUTE  = 0;
PRISMATIC = 1;

% Declare the symbolic variables
syms a11 a12 a13 a14 a21 a22 a23 a24 a31 a32 a33 a34 a41 a42 a43 a44
syms Theta
syms sx sy sz t
syms sox soy soz

sx = val_sx;
sy = val_sy;
sz = val_sz;
sox = val_sox;
soy = val_soy;
soz = val_soz;

Theta = 0;

Theta = symTheta;

if joint_type == REVOLUTE
    t = 0;
    % angle is not zero
elseif joint_type == PRISMATIC
    % Theta is 0 then Sin(0) = 0 and Cos(0) = 1
    Theta = 0;
    % t is not zero
else
    error('Joint type is wrong! REVOLUTE = 0 and PRISMATIC = 1');
end

% Formula according to 2.132
% Rotation matrix
a11 = (sx^2 - 1) * (1 - cos(Theta)) + 1;
a12 = sx * sy * (1 - cos(Theta)) - (sz * sin(Theta));
a13 = sx * sz * (1 - cos(Theta)) + (sy * sin(Theta));

a21 = sy * sx * (1 - cos(Theta)) + (sz * sin(Theta));
a22 = (sy^2 - 1) * (1 - cos(Theta)) + 1;
a23 = sy * sz * (1 - cos(Theta)) - (sx * sin(Theta));

a31 = sz * sx * (1 - cos(Theta)) - (sy * sin(Theta));
a32 = sz * sy * (1 - cos(Theta)) + (sx * sin(Theta));
a33 = (sz^2 - 1) * (1 - cos(Theta)) + 1;

a41 = 0;
a42 = 0;
a43 = 0;

% Translation Matrix
a14 = (t * sx) - (sox * (a11 - 1)) - (soy * a12) - (soz * a13);
a24 = (t * sy) - (sox * a21) - (soy * (a22 - 1)) - (soz * a23);
a34 = (t * sz) - (sox * a31) - (soy * a32) - (soz * (a33 - 1));
a44 = 1;

% Transform matrix
output = [a11 a12 a13 a14; a21 a22 a23 a24; a31 a32 a33 a34; a41 a42 a43 a44];