Updated axis limit computation and added other common trajectories. #1

juanmed · 2019-01-17T10:10:54Z

Hi. These are not big changes:

Sometimes the desired trajectory is outside the drawing limits. So I updated the set_limit calls to draw plot with appropriate limits.
Added polynomial and leminiscata (figure 8)trajectories since are common when analyzing performance of control algorithms. These trajectories are k-differentiable so they are useful for computing reference positions, velocities and accelerations in a trajectory (because are 3-differentiable).

…liminiscata trajectories.

juanmed · 2019-01-17T18:54:00Z

I just added an LQR Controller. Its design was based on a linearization of the system dynamics. Right now it is tuned for output performance rather than input energy performance, so it is a bit crazy.

hbd730 · 2019-01-18T02:42:00Z

Hi, looks like you have done a lot of work, well done! I have a quick run locally and realize the drone flies off track in a second, is that what you seen on your side? I will need to have a closer look later on when I get time.

juanmed · 2019-01-18T08:01:19Z

Hi, thank you for creating this simulator. The behaviour you describe is what I see. However this behaviour is dependent on both the trajectory and the control_frequency variable. Also it is not limited to my LQR controller but also to the PID controller already present in the repository shows this behaviour.

I would like to mention that I have not yet fully analyzed your code. Since this behaviour is dependent on the trajectory and control_frequency, I would like to first understand how your code works and then try to identify this behaviour as related to the simulator or the controller.

hbd730 · 2019-01-18T11:39:44Z

Em...looks like it is broken somehow, a few people report it before as well. I will look into it this weekend. I actually have a single threaded faster version in bitbucket, which I think does not have this issue. It removes scheduler which has problem.It does speed up help control loop. I will migrate those changes onto within this weekend, and could you retest and adapt your pull request once done? I will let you know when it is done. I really appreciate your effort, and hope we can work something out together. Regards Peter

…

On Fri., 18 Jan. 2019, 7:01 pm Juan, ***@***.***> wrote: Hi, thank you for creating this simulator. The behaviour you describe is what I see. However this behaviour is dependent on both the trajectory and the control_frequency variable. Also it is not limited to my LQR controller but also to the PID controller already present in the repository. I would like to mention that I have not yet fully analyzed your code. Since this behaviour is dependent on the trajectory and control_frequency, I would like to first understand how your code works and then try to identify this behaviour as related to the simulator or the controller. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#1 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AHd0qQKpcfTRsZnimHreOa45bKjStNMXks5vEX9PgaJpZM4aE3Da> .

juanmed · 2019-01-18T13:00:52Z

Yes, for sure. It would be nice your implementation of the thread. I found your small simulator flexible enough to make superficial and deep changes with ease, it has a nice graphical interface and the code easy to understand.. I would like to keep collaborating on it while keeping this characteristics.

hbd730 · 2019-01-18T23:07:22Z

Hi Juan,

I have push my threading change to master, and created a branch for you dev/lqr-controller, do you want to adapt your changes on that branch. Do you want me to add you as a contributor? Let me know your email.

hbd730 · 2019-01-18T23:11:46Z

.gitignore

@@ -0,0 +1,6 @@
+*.pyc


you can remove this now, I already add .gitignore

hbd730 · 2019-01-18T23:38:17Z

controller.py

+    des_x_dot, des_y_dot, des_z_dot = des_state.vel
+    des_x_ddot, des_y_ddot, des_z_ddot = des_state.acc
+    des_psi = des_state.yaw
+    des_psi_dot = des_state.yawdot    


it does not look like these are used, please review the input parameters, and only takes the one that is used.

hbd730 · 2019-01-18T23:45:13Z

controller.py

    return F, M
+
+def run_LQR(quad, des_state):


If you can, could you refactor this controller.py file a bit so it only includes control algoriuthm, and seperate out calibration parameters.

pid() - move pid specific parameters into a seperate file.

lqr() - could you refactor a bit so lqr specific matrix constant goes into a seperate file similar to model.params

we should create a folder called "control" and only includes controller.py and params for now.

hbd730 · 2019-01-18T23:49:49Z

controller.py

+    #
+    # In general    u = Nu*r - K(x -Nx*r) = -K*x + (Nu + K*Nx)*r
+    # where K are the LQR gains, Nu and Nx are reference input matrices.
+    # See more at p.493 "Feedback Control of Dynamic Systems, Franklink G,


Good summery, could you write some simple description on how you tune and design lqr matrix, and their effect on the control in general?

hbd730 · 2019-01-18T23:50:58Z

runsim.py

@@ -24,26 +24,34 @@ def render(quad):

 def attitudeControl(quad, time, waypoints, coeff_x, coeff_y, coeff_z):
    desired_state = trajGen3D.generate_trajectory(time[0], 1.2, waypoints, coeff_x, coeff_y, coeff_z)
-    F, M = controller.run(quad, desired_state)


please rename original run to run_pid() as mentioned above

hbd730 · 2019-01-18T23:55:32Z

quadPlot.py

@@ -15,15 +15,18 @@
 history = np.zeros((500,3))
 count = 0

-def plot_quad_3d(waypoints, args=()):
+def plot_quad_3d(waypoints, init_pos, args=()):


see if you can adapt your change to my thread-less part, I will review this part later on. I suggest you seperate this review into two, one is general clean up and quadPlot related change, the other is lqr controller. You can do this in your branch now and submit incremental change, which it is easy for me to review.

hbd730 · 2019-01-18T23:58:38Z

trajGen3D.py

+
+    return np.stack((x, y, z), axis=-1)
+
+def get_poly_waypoints(t,n):


Since your get_leminiscata_waypoints does not seem to be used yet, I think we should create a folder call trajectories, and move all trajectories generation files into it and create a new one for your work and submit that change as a review.
Make sure you write some unit test in the test folder :)

hbd730

Good work so far and hope you can see my comment and continue your work as suggested.

hbd730 · 2019-01-22T10:09:12Z

Hi Juan, Are you still interested in working on your pull request? I am quite happy to work together and hope that your change can get in. Regards Peter

…

On Sat, 19 Jan 2019 at 00:00, Juan ***@***.***> wrote: Yes, for sure. It would be nice your implementation of the thread. I found your small simulator flexible enough to make superficial and deep changes with ease, it has a nice graphical interface and the code easy to understand.. I would like to keep collaborating on it while keeping this characteristics. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#1 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AHd0qaJPYq5YAbIKU2m3kGBAZ4ctX-aQks5vEcWFgaJpZM4aE3Da> .

juanmed · 2019-01-22T11:20:18Z

Hey! Yes of course. I have been busy with other stuff so I have not been able to update the branch. Actually I have been working on a differential flatness based controller. As time allows I will update both the LQR controller and try to merge the differential flatness based controller.

By the way, I was wondering if it is possible for you to add some comments in your code. For example, adding the euler convention used and other details. I think it might make your code easier to undertand.

Thanks for updating the repository!

hbd730 · 2019-08-09T12:35:50Z

Hi Juan, How are you! I recently receive a few email queries about my simulator, people are seeing quadcopter flying away, which never happen to me before until recently I download it myself and have a look. I can no longer get it fly like it did before on another PC, please see here https://www.youtube.com/watch?v=jlmXEo9Jmxc. The quad-rotor does fly away now... I vaguely remember you also encountered this issue before, and I saw your fork, and you did pretty good job on extending my work. I think you are a expert now :) I am wondering if you have a fix for yaw control for helix trajectory. If you do have, could you please share so it could help other people as well. Best Regards Peter

…

On Fri, 9 Aug 2019 at 22:32, Peter Huang ***@***.***> wrote: Hi Juan, How are you! I recently receive a few email queries about my simulator, people are seeing quadcopter flying away, which never happen to me before until recently I download it myself and have a look. I can no longer get it fly like it did before on another PC, please see the attachment. The quad-rotor does fly away... I vaguely remember you also encountered this issue before, and I saw your fork, and you did pretty good job on extending my work, you are a expert now :) I am wondering if you have a fix for yaw control for helix trajectory. If you do have, could you please share so it could help other people as well. Best Regards Peter On Tue, 22 Jan 2019 at 22:20, Juan ***@***.***> wrote: > Hey! Yes of course. I have been busy with other stuff so I have not been > able to update the branch. Actually I have been working on a differential > flatness based controller. As time allows I will update both the LQR > controller and try to merge the differential flatness based controller. > > By the way, I was wondering if it is possible for you to add some > comments in your code. For example, adding the euler convention used and > other details. I think it might make your code easier to undertand. > > Thanks for updating the repository! > > — > You are receiving this because you commented. > Reply to this email directly, view it on GitHub > <#1 (comment)>, > or mute the thread > <https://github.com/notifications/unsubscribe-auth/AHd0qc53v7G_bA2bcaREZTaaf571KnWNks5vFvPygaJpZM4aE3Da> > . >

juanmed · 2019-08-15T09:47:09Z

Hey Peter,

Thanks for your message. You can find my changes regarding heading here on the trajGen3D.py file. But I cannot guarantee that these changes will not make the drone to fly away since I did not directly tried to solve that. However I should mention that this fix plus update you pushed from bitbucked helped. I did not see the drone fly away anymore (of course, if the control inputs are correct).

juanmed added 4 commits January 17, 2019 18:52

Updated axis limits to draw for any trajectory. Added polynomial and …

2d265fe

…liminiscata trajectories.

Deleted .gif

c870974

Removed unnecessary files.

d98acab

Added an LQR Controller in controller.py

b72ff9b

hbd730 reviewed Jan 18, 2019

View reviewed changes

.gitignore

@@ -0,0 +1,6 @@

*.pyc

Copy link

Owner

hbd730 Jan 18, 2019

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

you can remove this now, I already add .gitignore

hbd730 reviewed Jan 18, 2019

View reviewed changes

hbd730 reviewed Jan 19, 2019

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Updated axis limit computation and added other common trajectories. #1

Updated axis limit computation and added other common trajectories. #1

juanmed commented Jan 17, 2019 •

edited

Loading

juanmed commented Jan 17, 2019

hbd730 commented Jan 18, 2019

juanmed commented Jan 18, 2019 •

edited

Loading

hbd730 commented Jan 18, 2019 via email

juanmed commented Jan 18, 2019

hbd730 commented Jan 18, 2019 •

edited

Loading

hbd730 Jan 18, 2019

hbd730 Jan 18, 2019

hbd730 Jan 18, 2019

hbd730 Jan 18, 2019

hbd730 Jan 18, 2019

hbd730 Jan 18, 2019

hbd730 Jan 18, 2019

hbd730 left a comment

hbd730 commented Jan 22, 2019 via email

juanmed commented Jan 22, 2019

hbd730 commented Aug 9, 2019 via email

juanmed commented Aug 15, 2019


		return np.stack((x, y, z), axis=-1)

		def get_poly_waypoints(t,n):

Updated axis limit computation and added other common trajectories. #1

Are you sure you want to change the base?

Updated axis limit computation and added other common trajectories. #1

Conversation

juanmed commented Jan 17, 2019 • edited Loading

juanmed commented Jan 17, 2019

hbd730 commented Jan 18, 2019

juanmed commented Jan 18, 2019 • edited Loading

hbd730 commented Jan 18, 2019 via email

juanmed commented Jan 18, 2019

hbd730 commented Jan 18, 2019 • edited Loading

hbd730 Jan 18, 2019

Choose a reason for hiding this comment

hbd730 Jan 18, 2019

Choose a reason for hiding this comment

hbd730 Jan 18, 2019

Choose a reason for hiding this comment

hbd730 Jan 18, 2019

Choose a reason for hiding this comment

hbd730 Jan 18, 2019

Choose a reason for hiding this comment

hbd730 Jan 18, 2019

Choose a reason for hiding this comment

hbd730 Jan 18, 2019

Choose a reason for hiding this comment

hbd730 left a comment

Choose a reason for hiding this comment

hbd730 commented Jan 22, 2019 via email

juanmed commented Jan 22, 2019

hbd730 commented Aug 9, 2019 via email

juanmed commented Aug 15, 2019

juanmed commented Jan 17, 2019 •

edited

Loading

juanmed commented Jan 18, 2019 •

edited

Loading

hbd730 commented Jan 18, 2019 •

edited

Loading