CentroidalMPC.advance() gets stuck with Ipopt and Mumps #900
Comments
|
Do collect a bit more information, probably you can attach a debugger to the Python process and get a backtrace of the thread that is hanging, to get a bit more context of what is happening. What is the usage of CPU cores? Can you report the full conda environment you are using? Can you also try to set |
|
Also conda-forge/mumps-feedstock#126 seems related. |
|
Thanks a lot @traversaro !! Setting
When the code hangs, the CPU usage is 0%, so a deadlock is likely the issue.
I'm using pixi, below you can find my environment: pixi list
|
The
|
|
By the way, I tried to reproduce the issue with https://raw.githubusercontent.com/ami-iit/bipedal-locomotion-framework/da3d948990fdab29daf8da9ccb90fb93a093804b/bindings/python/ReducedModelControllers/tests/test_centroidal_mpc.py failing, so if you can provide an easy way to reproduce the issue it would be great. |
Here is the [project]
authors = ["Carlotta Sartore <[email protected]>"]
channels = ["conda-forge"]
description = "Suite of parametrized controller with simulation environments for co-design of Humanoid Robots"
name = "comodo"
platforms = ["linux-64"]
version = "0.0.0"
repository = "https://github.com/ami-iit/comodo"
readme = "README.md"
license-file = "LICENSE"
[activation]
# To avoid issue with git-lfs, see https://github.com/prefix-dev/pixi/issues/2000
env = { GIT_LFS_SKIP_SMUDGE = "1" }
[dependencies]
# bipedal-locomotion-framework and manifpy have pinned build dependencies
# due to https://github.com/ami-iit/bipedal-locomotion-framework/issues/885
bipedal-locomotion-framework = { version = "==0.18.0", build = "ha770c72_15" }
manifpy = { version = "==0.0.5", build = "py310hf5d888b_0" }
idyntree = "*"
adam-robotics = "*"
mujoco = "*"
mujoco-python-viewer = "*"
matplotlib = ">=3.9.2,<4"
urllib3 = "*"
urchin = "*"
resolve-robotics-uri-py = "*"
notebook = "*"
rod = "*"
ipykernel = "*"
mesalib = "*"
optuna = ">=4.0.0,<5"
cmaes = ">=0.11.1,<0.12"
plotly = ">=5.24.1,<6"
casadi = "<=3.6.6"
nbformat = ">=5.10.4,<6"
python-kaleido = ">=0.2.1,<0.3"
scikit-learn = ">=1.5.2,<2"
# jaxsim = ">=0.4.2"
[pypi-dependencies]
jaxsim = { git = "https://github.com/ami-iit/jaxsim.git", rev="main", extras = ["all"]}
# jaxsim = { path = "../jaxsim", extras = ["all"] }
urdfmodifiers = { git = "https://github.com/CarlottaSartore/urdf-modifiers.git", rev = "scalar_modification" }
comodo = { path = "./", editable = true }And here the output of |
Sorry, this was not clear perhaps. The reproduction attempt was "failing" as the test was passing and so I "failed" in reproducing the issue. |
Sure, this is the script I'm testing right now: https://github.com/xela-95/comodo/blob/viscoelastic-simulations/examples/tune_jaxsim_contact_model.py In order to reproduce the issue just comment out https://github.com/xela-95/comodo/blob/8510ead3e1f3c3361c3852134dfeeb39025bdd2a/examples/tune_jaxsim_contact_model.py#L29 and run it. I usually run it with: |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Hi, I'm using comodo to perform simulations on JaxSim, since it provides an easy to use interface to blf controllers.
However I'm having some flaky errors (not repeating 100% of the times) in which the BipedalLocomotion::ReducedModelControllers::CentroidalMPC::advance method gets stuck and do not exit anymore (even with ctrl-C). When i go debugging those methods it just logs:
That method is called in
comodowrapper of the centroidal MPCcentroidalMPC.plan_trajectory(): https://github.com/ami-iit/comodo/blob/1130009d6ef0cfa1944495209a15887388faa52d/src/comodo/centroidalMPC/centroidalMPC.py#L36-L57Have you ever encountered this kind of issue before? What should I try to change?
System info
CC @CarlottaSartore
The text was updated successfully, but these errors were encountered: