| Both sides previous revisionPrevious revisionNext revision | Previous revision |
| ece4560:piktul:05resratepos [2017/10/22 16:24] – pvela | ece4560:piktul:05resratepos [2024/08/20 21:38] (current) – external edit 127.0.0.1 |
|---|
| end | end |
| </code> | </code> |
| This can be a hidden function within the ''genPositionTrajectory'' member function or it can be a protected member function (the easiest would be to make it a public member function). It is the main part that has the resolved rate equations. Code stubs to insert into the piktul class can be found here: {{ece4560:piktul:resolvedStubs.m|stubs}}. | This can be a hidden function within the ''genPositionTrajectory'' member function or it can be a protected member function (the easiest would be to make it a public member function). It is the main part that has the resolved rate equations. Code stubs to insert into the piktul class can be found here: {{ece4560:piktul:resolvedStubs.m|stubs}}. The code stubs have ''posODE'' as a hidden function within a function. The documentation of the code stub provides additional clues as to how to code the entire learning module up. |
| |
| Suppose you'd like to test it out without actually using the piktul, but rather you'd like to use the ''piktul_display'' function. The easiest would be to write a modified version of the ''followJointTrajectory'' function called ''displayJointTrajectory'' | Overall, try to have functionality or code that seems inherent to the piktul actually reside in the ''piktul'' class. Some can be outside of it. I leave this as a judgement call, but definitely you should be considering what functionality should actually be implemented within the class and how it supports abstraction/encapsulation as per good coding design. |
| | |
| | |
| | |
| | |
| | ==== Playing Back as a Movie ==== |
| | |
| | Suppose you'd like to test it out without actually using the piktul. You can always use the ''piktul_display'' function to test out the code. If you can get the resolved rate code working to |
| | generate the joint angles versus time, then using ''piktul_display'' in a loop will display it as a movie in a Matlab figure. The easiest would be to write a modified version of the ''followJointTrajectory'' function called ''displayJointTrajectory'' |
| <code=Matlab> | <code=Matlab> |
| function displayJointTrajectory(jtraj, tspan, nsteps) | function displayJointTrajectory(jtraj, tspan, nsteps) |
| end | end |
| </code> | </code> |
| that uses the function ''piktul_display'' instead of ''setArm''. doing so will visualize the arm instead of commanding the arm. | that uses the function ''piktul_display'' instead of ''setArm''. Doing so will visualize the arm instead of commanding the arm. Remember to include the ''drawnow'' command at the end of the loop to force the figure to refresh. |
| | ==== Playing Back as a Command Sequence ==== |
| ==== Playing Back as a Movie/Command Sequence ==== | |
| |
| For this version, you will have to actually use the optional time argument to ''setArm'' which is the time duration of the movement. Whatever time step you use in the ''for'' loop to move the arm, you will have to send a time duration that is 10\% longer, then pause for the actual duration before resuming the next loop iteration. The reason being that the little controller board actually implements spline smoothing to go from one joint angle to another, so if you specify the | For this version, you will have to actually use the optional time argument to ''setArm'' which is the time duration of the movement. Whatever time step you use in the ''for'' loop to move the arm, you will have to send a time duration that is 10\% longer, then pause for the actual duration before resuming the next loop iteration. The reason being that the little controller board actually implements spline smoothing to go from one joint angle to another, so if you specify the |
| |
| Be careful about the whole radians versus degrees thing. The ''piktul'' inputs are in degrees, so you'll probably need to use Matlab's ''cosd'' and ''sind'' functions, which take the angular arguments in degrees. Also, the angular velocities will have to be scaled by $180/\pi$ to convert them to degrees since the pseudo-inverse of the Jacobian will give results that are sensible in radians per second (not degrees per second). Within the position Jacobian function, the conversion to degrees for the output should already be done. | Be careful about the whole radians versus degrees thing. The ''piktul'' inputs are in degrees, so you'll probably need to use Matlab's ''cosd'' and ''sind'' functions, which take the angular arguments in degrees. Also, the angular velocities will have to be scaled by $180/\pi$ to convert them to degrees since the pseudo-inverse of the Jacobian will give results that are sensible in radians per second (not degrees per second). Within the position Jacobian function, the conversion to degrees for the output should already be done. |
| | |
| | ==== Creating the Position Trajectory ==== |
| | |
| | To make a constant function to pass as an argument in a |
| | trajectory structure, here is a simple example: |
| | <code> |
| | > posTraj.position = @(t) 2 + 5*t; |
| | > posTraj.velocity = @(t) 5; |
| | </code> |
| | Your version will have to be vector valued with the correct values. |
| | |
| |
| --------- | --------- |
