Added a example: crane moving a load #197

Peter230655 · 2024-07-27T11:56:18Z

A load hanging on the horizontal jib of a crane is to be moved from A to B as fast as possible. However, the load must not 'swing over' the final position B, say there is a wall it should not crash into. opty's solution is quite different from what I naively expected.

Peter230655 · 2024-07-27T12:08:03Z

I tried to copy all 'formalities' from an existing example in examples_gallery, so I hope it will run with sphinx.

Peter230655 · 2024-07-29T06:04:45Z

This seems to look o.k. in html

moorepants · 2024-07-30T04:46:49Z

Each plot should be associated with a new cell so that the plots are maximized in size and readable. These are not:

moorepants · 2024-07-30T04:47:55Z

I ran this and it failed:

message from optimizer: b"Restoration phase failed, algorithm doesn't know how to proceed."
Iterations needed 338
Objective value  2.338e-02

Examples should be robust and always converge. The most reliable way is to give a better initial guess (possibly even a solution).

moorepants · 2024-07-30T04:49:04Z

This seems to look o.k. in html

Yes, nice work!

moorepants · 2024-07-30T04:49:43Z

examples-gallery/plot_crane_moving_a_load.py

@@ -0,0 +1,294 @@
+# %%
+"""
+Crane moving a load


Suggested change

Crane moving a load

Crane Moving a Load

moorepants · 2024-07-30T04:50:31Z

examples-gallery/plot_crane_moving_a_load.py

+
+A load is moved by a crane. The load is rotably connected to the crane by a
+massless rod. The goal is to move the load from the initial position to the
+final position in the shortest possible time. The load must not 'overswing'


Suggested change

final position in the shortest possible time. The load must not 'overswing'

final position in the shortest possible time. The load must not over-swing

moorepants · 2024-07-30T04:51:42Z

examples-gallery/plot_crane_moving_a_load.py

+**Constants**
+
+- l: length of the rod attaching the load to the crane [m]
+- m1: mass of mover attached to the arm of the crane [kg]


The variables should be latex or code:

- :math:`m_1` OR ``m1``

The variables should be latex or code:

- :math:`m_1` OR ``m1``

I had played around with :math:..... but did not get it to work right.
Will it only work in lines like:

"""

:math: ... .

"""
not in lines starting with # ?

Math in RestructuredText is explained here:

https://ikerdocs-sphinx.readthedocs.io/syntax/math.html

Math in RestructuredText is explained here:

https://ikerdocs-sphinx.readthedocs.io/syntax/math.html

This is what I looked at, but never got it to work. I will play around some more.
As I understand, :math: or ..math: only works in docstrings. In simulations for examples_gallery there is a docstring at the very beginning.
Question: are docstrings allowed further down in the text, or to be avoided?

Sphinx-gallery treats the module level "docstring" as a "cell" of RestructuredText.

This has not bee addressed.

moorepants · 2024-07-30T04:52:55Z

examples-gallery/plot_crane_moving_a_load.py

+- uxc: velocity of the mover in x-direction [m/s]
+- uxl: velocity of the load in x-direction [m/s] dependent speed
+- uyl: velocity of the load in y-direction [m/s] dependent speed
+- u: angular velocity of the rod [rad/s]


The states are technically only the minimal set of coordinates and speeds, which are likely q and u.

The states are technically only the minimal set of coordinates and speeds, which are likely q and u.

I did not know this definition, I will correct this and the other suggestions. I needed the dependent coordinates to model the non over . swing.

I see below you are doing a non-minimal coordinate formulation. The states are still then only the independent coordinates and speeds.

The states are technically only the minimal set of coordinates and speeds, which are likely q and u.

they are xc, uxc, q, u.

Peter230655 · 2024-07-30T04:52:58Z

Each plot should be associated with a new cell so that the plots are maximized in size and readable. These are not:

I know how cto fix it, fill come soon in the next commit.

moorepants · 2024-07-30T04:53:35Z

examples-gallery/plot_crane_moving_a_load.py

+
+"""
+import sympy.physics.mechanics as me
+from collections import OrderedDict


OrderedDict is no longer needed, it is a remnant of Python 2 not ordering dictionaries by default.

moorepants · 2024-07-30T04:56:32Z

examples-gallery/plot_crane_moving_a_load.py

+
+# %%
+# Set up Kane's equations of motion.
+# Just the standard way of doing it.


Remove the second line, unnecessary.

moorepants · 2024-07-30T04:57:05Z

examples-gallery/plot_crane_moving_a_load.py

+P1a = me.Particle('P1a', P1, m1)
+P2a = me.Particle('P2a', P2, m2)
+
+BODY = [P1a, P2a]


Suggested change

BODY = [P1a, P2a]

bodies = [P1a, P2a]

Use PEP8 naming conventions if at all possible.

Peter230655 · 2024-07-30T04:57:20Z

I ran this and it failed:
message from optimizer: b"Restoration phase failed, algorithm doesn't know how to proceed."
Iterations needed 338
Objective value  2.338e-02
Examples should be robust and always converge. The most reliable way is to give a better initial guess (possibly even a solution).

Strange. I never happened with me.

moorepants · 2024-07-30T04:57:21Z

examples-gallery/plot_crane_moving_a_load.py

+
+BODY = [P1a, P2a]
+
+FL = [(P1, F * N.x - m1*g*N.y), (P2, -m2*g*N.y)]


Suggested change

FL = [(P1, F * N.x - m1*g*N.y), (P2, -m2*g*N.y)]

forces = [(P1, F * N.x - m1*g*N.y), (P2, -m2*g*N.y)]

moorepants · 2024-07-30T04:59:16Z

examples-gallery/plot_crane_moving_a_load.py

+    configuration_constraints=config_constr,
+    velocity_constraints=speed_constr)
+
+(fr, frstar) = KM.kanes_equations(BODY, FL)


Suggested change

(fr, frstar) = KM.kanes_equations(BODY, FL)

fr, frstar = KM.kanes_equations(bodies, loads=forces)

moorepants · 2024-07-30T04:59:43Z

examples-gallery/plot_crane_moving_a_load.py

+    velocity_constraints=speed_constr)
+
+(fr, frstar) = KM.kanes_equations(BODY, FL)
+EOM = kd.col_join(fr + frstar)


Suggested change

EOM = kd.col_join(fr + frstar)

eom = kd.col_join(fr + frstar)

moorepants · 2024-07-30T05:00:33Z

examples-gallery/plot_crane_moving_a_load.py

+# Set up the optimization problem and solve it.
+# Note: While opty could calculate the reaction forces on, say, the mover,
+# (add them to the state variables), my trials showed that even with this
+# small example the calculation time increases enourmously.


The note is not helpful in this example, as it is vaguely explained and likely just distracts. I recommend removing it.

moorepants · 2024-07-30T05:03:26Z

examples-gallery/plot_crane_moving_a_load.py

+state_symbols = tuple((*q_ind, *q_dep, *u_ind, *u_dep))
+constant_symbols = (l, m1, m2, g)
+specified_symbols = (F,)
+h_opty = sm.symbols('h_opty')


I'm curions why you've chosen to use h_opty. I changed the variable name internally to this so that there would be no clashes with using h as a symbol name by the user. It is my intention that a user would never know about this naming convention for the internals and showing it here reveals that!

I'm curions why you've chosen to use h_opty. I changed the variable name internally to this so that there would be no clashes with using h as a symbol name by the user. It is my intention that a user would never know about this naming convention for the internals and showing it here reveals that!

Now I understand your intention with h_opty. Will change back to h.

moorepants · 2024-07-30T05:03:52Z

examples-gallery/plot_crane_moving_a_load.py

+interval_value = h_opty
+
+# Specify the known system parameters.
+par_map = OrderedDict()


Suggested change

par_map = OrderedDict()

par_map = {}

I had wondered about the OrderedDict - but just copied it from your examples. Will change it.

moorepants · 2024-07-30T05:04:27Z

examples-gallery/plot_crane_moving_a_load.py

+    grad[-1] = 1.
+    return grad
+
+# start location and and final location of the load.


Make a cell so text is rendered.

moorepants · 2024-07-30T05:05:16Z

examples-gallery/plot_crane_moving_a_load.py

+
+# start location and and final location of the load.
+anfang = 0.0
+ende = 15.0


It is preferable if the documentation is all in English to be consistent.

It is preferable if the documentation is all in English to be consistent.

I did this because in my early python times, I used English words, which were standard Python terms - and then I wondered why my program did not run (like I called a variable 'range'. Will change it to English.

That's a good method, but we should just make the docs consistent in language. Sphinx does have mechanisms to make multiple versions of the docs in different languages. I've never done that, but I'm sure it would be helpful to many learners.

moorepants · 2024-07-30T05:06:28Z

examples-gallery/plot_crane_moving_a_load.py

+constant_symbols = (l, m1, m2, g)
+specified_symbols = (F,)
+h_opty = sm.symbols('h_opty')
+unknown_symbols = []


Is this used below?

Is this used below?

No. I also copied from some example, maybe from Timo. Allows the prob = Problem(..) to be always the same.
Better delete it?

If it isn't used, then delete it.

moorepants · 2024-07-30T05:07:09Z

examples-gallery/plot_crane_moving_a_load.py

+    )
+# %%
+# Using a 'reasonable' initial guess speeds up the optimization process a lot.
+# 'Bad' initial guesses may even prevent convergence.


Just state what initial guess you are constructing. You can plot it easily with plot_trajectories() to show the reader what it is.

moorepants · 2024-07-30T05:08:09Z

examples-gallery/plot_crane_moving_a_load.py

+
+
+# this allows to change the maximum number of iterations.
+# Standard is 3000.


Still a code comment. Render as cell text.

moorepants · 2024-07-30T05:08:57Z

examples-gallery/plot_crane_moving_a_load.py

+# Standard is 3000.
+prob.add_option('max_iter', 1000)
+# Find the optimal solution.
+for _ in range(1):


Remove this loop if there is only one iteration.

Ideally all examples solve with a single call to .solve().

moorepants · 2024-07-30T05:09:10Z

examples-gallery/plot_crane_moving_a_load.py

+print(f"Objective value {solution[-1]: .3e}")
+
+# %%
+# plot the accuracy of the solution.


Suggested change

# plot the accuracy of the solution.

# Plot the accuracy of the solution.

Peter230655 · 2024-08-03T14:47:18Z

I changed, say, q to q, but it seems I cannot use `` in commit comments.

moorepants · 2024-08-03T16:15:31Z

I made a number of formatting fixes and tried to correct the fps. You can see what I changed here:

082d9eb

You can do git pull origin plot_crane_moving_a_load to get my updates in your branch.

Two issues remain:

One out of three times this fails to find an optimal solution for me. If the initial guess can be improved, maybe that will not happen.
I don't think the animation is animating to the final time, as it doesn't look like it ever hits the wall. This may be due to the way you are generating the animation frames and that arange() will skip the actual last moment in time when it hits the wall.

Peter230655 · 2024-08-05T15:05:31Z

I made a number of formatting fixes and tried to correct the fps. You can see what I changed here:
I finally understood! Embarrassing how long it takes me at times!

082d9eb

You can do git pull origin plot_crane_moving_a_load to get my updates in your branch.

Two issues remain:
1. One out of three times this fails to find an optimal solution for me. If the initial guess can be improved, maybe that will not happen.

2. I don't think the animation is animating to the final time, as it doesn't look like it ever hits the wall. This may be due to the way you are generating the animation frames and that `arange()` will skip the actual last moment in time when it hits the wall.

I improved the initial guess by using the solution of a previous run. It never failed to find a solution with me, at least always an acceptable one. What I pushed found an optimal solution with me.
I thought arange would get the last point. my test showed: The last point of np.arange(...) is 5.85, while the real end = (num_nodes - 1) * h = 5.8541. Could this small difference account for what you are seeing?
I just got this error from Github:

This is the file where I store the solution as initial guess for later runs.. I had no problem with it doing make html ?
What I mean is this:
I add the file in my branch.
Then I switch to the Anaconda Prompt and I give the command make html.
After this is finished, I go to docs in my explorer, and search for index.html
There are three of them and the 3rd one is correct. It shows the crane correctly.

Is this Github error a fatal one, so that my simulation cannot be merged like this? Then of course I will change it.
Thanks!

Peter230655 · 2024-08-24T05:37:30Z

In this example, I stored the result of a run, and use this result like this: initial_guess = np.load('solution of a previous run').
This makes it a bit faster.
Is this acceptable for examples-gallery? Will examples-gallery 'have' this previous solution?

I think, the failing test is because Github does not 'have' the previous solution ?

moorepants · 2024-09-05T06:26:43Z

examples-gallery/plot_crane_moving_a_load.py

+# - i4 = [-par_map[l] for _ in range(num_nodes)]
+# - i5 = [0.0 for _ in range(5*num_nodes)]
+# - i6 = [0.01]
+# - initial_guess = np.array(i1 + i2 + i3 + i4 + i5 + i6)


Move this to a full comment cell as an explanation, something like:

# %% # This initial guess was used to get a solution which forms the actual initial # guess:: # # i1 = ... # i2 = ...

moorepants · 2024-09-05T06:28:13Z

examples-gallery/plot_crane_moving_a_load.py

+
+anim = animation.FuncAnimation(fig, update,
+                               frames=np.arange(t0, tf, 1/fps),
+                               interval=1/fps*1000)


The animation does not run until it hits the for me. Maybe it is running out of memory. I suggest reduce the fps and number of frames.

moorepants · 2024-09-05T06:58:20Z

examples-gallery/plot_crane_moving_a_load.py

-#        range(num_nodes)]
+# guess::
+#
+# - i1 = [(ending_location - starting_location)/num_nodes*i for i in range(num_nodes)]


The spacing needs to be such that it renders as code.

https://www.sphinx-doc.org/en/master/usage/restructuredtext/basics.html#literal-blocks

The spacing needs to be such that it renders as code.

I tried to go as per the sphinx instructions you sent.

Your idea to execute it but then just overwrite the initial guess with the solution is better.

Your idea to execute it but then just overwrite the initial guess with the solution is better.

I will change this and make a new PR.

Please don't make new prs for the same topic, just update this one.

Please don't make new prs for the same topic, just update this one.

Sorry, I meant I would add to this one only!

moorepants · 2024-09-05T16:22:32Z

examples-gallery/plot_crane_moving_a_load.py

+# 'solution.npy', to speed up convergence. To get this solution, the initial
+# guess above was used, which is now overwritten.
+
+initial_guess = np.load('solution.npy')


Can you rename the file to crane_moving_a_load_solution.npy because we will likely add more of these files. I think even another PR has one named the same and they are all in the same directory.

Can you rename the file to crane_moving_a_load_solution.npy because we will likely add more of these files. I think even another PR has one named the same and they are all in the same directory.

Makes emminent sense! Will do so right now

moorepants · 2024-09-05T16:45:41Z

You need to git rm solution.npy.

Peter230655 · 2024-09-05T16:47:36Z

You need to git rm solution.npy.

..and then push again? (By the time I will be 75 I'll be o.k. :-) )

moorepants · 2024-09-05T16:48:10Z

Yes

Peter230655 · 2024-09-05T16:49:38Z

Yes

Can I push while the old check is still running?

moorepants · 2024-09-05T16:54:00Z

Yes

moorepants · 2024-09-05T17:08:15Z

examples-gallery/plot_crane_moving_a_load.py

+
+# %%
+# The initial guess is the solution of some previous run, saved in the file
+# 'solution.npy', to speed up convergence. To get this solution, the initial


Needs updating to new filename.

Needs updating to new filename.

How come you see EVERYTHING?!? I guess, because you are a teacher by trade. Will change it right now.

Yes, maybe grading papers causes that.

moorepants · 2024-09-05T17:40:57Z

Thanks! Nice example.

Peter230655 · 2024-09-06T05:40:12Z

Thanks! Nice example.

Thank YOU! I must admit, it makes me feel good! :-)
What I like about this example: Intuitively I had expected that it would crawl towards the right wall. Actually, it goes full speed!
opty often (in my toy eamples) finds an unexpected (by me) solution.
NB: I am quite pleased that I found a solution to convert the variable h into a correctly rounded number for plot_constraint_violations. :-)

1. commit of crane_moving_a_load

e3ba35e

2. commit

c05f1b5

moorepants changed the title ~~1. commit of crane_moving_a_load~~ Added a example: crane moving a load Jul 30, 2024

moorepants reviewed Jul 30, 2024

View reviewed changes

changed the statrs from, say, q to q

980fdbc

Various formatting fixes and fps fix.

082d9eb

changes by JM incorportated, use a previous solution as initial guess

bd290d2

Peter230655 added 2 commits August 5, 2024 17:46

reduced num_nodes, now takes 22.5 sec to run

5087a5f

corrections as per JM's suggestions for the steamboat

999c8bb

now with the .npy file

85b5b75

moorepants added the example label Aug 30, 2024

moorepants reviewed Sep 5, 2024

View reviewed changes

set fps = 10 (before fps = 20). Improved print of initial guess used

4f67c5c

moorepants reviewed Sep 5, 2024

View reviewed changes

calculate initial guess, then overwrite it

c856f0d

moorepants reviewed Sep 5, 2024

View reviewed changes

renamed the file holding the solution

4ad626f

removed old solution file

c4f9151

moorepants reviewed Sep 5, 2024

View reviewed changes

renamed the solution file in the description

4f4950d

moorepants merged commit 276b08d into csu-hmc:master Sep 5, 2024
21 checks passed

Peter230655 deleted the plot_crane_moving_a_load branch September 6, 2024 05:42

	final position in the shortest possible time. The load must not 'overswing'
	final position in the shortest possible time. The load must not over-swing


		BODY = [P1a, P2a]

		FL = [(P1, F * N.x - m1gN.y), (P2, -m2gN.y)]

	FL = [(P1, F * N.x - m1gN.y), (P2, -m2gN.y)]
	forces = [(P1, F * N.x - m1gN.y), (P2, -m2gN.y)]

	(fr, frstar) = KM.kanes_equations(BODY, FL)
	fr, frstar = KM.kanes_equations(bodies, loads=forces)

	EOM = kd.col_join(fr + frstar)
	eom = kd.col_join(fr + frstar)



		# this allows to change the maximum number of iterations.
		# Standard is 3000.

	# plot the accuracy of the solution.
	# Plot the accuracy of the solution.

Added a example: crane moving a load #197

Added a example: crane moving a load #197

Conversation

Peter230655 commented Jul 27, 2024

Peter230655 commented Jul 27, 2024

Peter230655 commented Jul 29, 2024

moorepants commented Jul 30, 2024

moorepants commented Jul 30, 2024

moorepants commented Jul 30, 2024

Choose a reason for hiding this comment

moorepants Jul 30, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Peter230655 Jul 30, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Peter230655 Jul 30, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Peter230655 Jul 30, 2024 • edited Loading

Choose a reason for hiding this comment

Peter230655 commented Jul 30, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Peter230655 commented Jul 30, 2024

Choose a reason for hiding this comment

moorepants Jul 30, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Peter230655 commented Aug 3, 2024

moorepants commented Aug 3, 2024

Peter230655 commented Aug 5, 2024 • edited Loading

Peter230655 commented Aug 24, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Peter230655 Sep 5, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

moorepants commented Sep 5, 2024

Peter230655 commented Sep 5, 2024

moorepants commented Sep 5, 2024

Peter230655 commented Sep 5, 2024

moorepants commented Sep 5, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

moorepants commented Sep 5, 2024

Peter230655 commented Sep 6, 2024

moorepants Jul 30, 2024 •

edited

Loading

Peter230655 Jul 30, 2024 •

edited

Loading

Peter230655 Jul 30, 2024 •

edited

Loading

Peter230655 Jul 30, 2024 •

edited

Loading

moorepants Jul 30, 2024 •

edited

Loading

Peter230655 commented Aug 5, 2024 •

edited

Loading

Peter230655 commented Aug 24, 2024 •

edited

Loading

Peter230655 Sep 5, 2024 •

edited

Loading