porting examples from tesseract_planning

[jf---]

Hi,

I'd like to contribute by porting the examples in the tesseract_planning repo to python.
I've started with

• glass_upright_example.cpp
• puzzle_piece_example.cpp

So far I can't get the solver to converge, for glass_upright_example.py I get the following:

[WARN] convex solver failed! set TRAJOPT_LOG_THRESH=DEBUG to see solver output. saving model to /tmp/fail.lp and IIS to /tmp/fail.ilp
ERROR in LDL_factor: Error in KKT matrix LDL factorization when computing the nonzero elements. The problem seems to be non-convex
ERROR in osqp_setup: KKT matrix factorization.
The problem seems to be non-convex.
OSQP Setup failed with error: Could not initialize OSQP: error 5
[WARN] convex solver failed! set TRAJOPT_LOG_THRESH=DEBUG to see solver output. saving model to /tmp/fail.lp and IIS to /tmp/fail.ilp
[ERROR] The convex solver failed you one too many times.

glass_upright_example.py doesn't give any warnings, but isn't successful in planning the path.
(I still need to rule out that parsing puzzle_bent.csv isn't the issue)

I've noticed that in the windows version of tesseract_python the QPOASES, QSQP solvers are available, but not GUROBI and BPMPD, while the latter IIRC is the default solver. Is it possible that this may be the culprint?

[Levi-Armstrong]

I am excited to see this.

[Levi-Armstrong]

The solver could be the issue. I am not sure if those examples via c++ use the default or explicitly set one.

[johnwason]

Thanks @jf--- . What platform are you running and how did you install the Python package?

[jf---]

@Levi-Armstrong thx, I'll keep chipping at it.

I am not sure if those examples via c++ use the default or explicitly set one

I don't see any solver set explicitly

The version I've installed doesn't come with IFOPT, so likely the default is set by TrajOptMotionPlannerTask in the YAML config file

What platform are you running and how did you install the Python package?

Windows 11 ( ARM, under protest, seriously, I'm working on a mac m1 ARM processor emulating, running windows in a VM, with windows emulating x86 🤦🏼 ), via pip, Version: 0.2.0.post1. The test suite runs just fine.

[jf---]

Sorry, I forgot to point out that pip install compas is required for now.
I missed or overlooked Vector3d class for cross and unitize methods

[johnwason]

@jf--- I think it should be possible to compile everything for native arm64 mac os. We haven’t made mac os or arm support a priority.

[jf---]

@johnwason it's very tempting to have proper mac support but I'm not particularly strong on cmake / c++. But with all the libs in conda it's probably doable. For now I'm probably one of very very few that has this need, so I'll endure VM's for now. I'd be happy to sponsor a Mac port.

FWIW, as I'm learning I'm trying to get to a pythonic interface on top of tesseract-python. Will report back when the code can withstand some daylight. For now having fun porting.

[johnwason]

Also take a look at https://github.com/rpiRobotics/rpi_general_robotics_toolbox_py/blob/master/src/general_robotics_toolbox/tesseract.py

[rjoomen]

I believe the default solver is OSQP nowadays.

[jf---]

Thanks for the pointer @johnwason thx for the heads-up @rjoomen

[dave992]

I am at the moment looking at porting the basic_cartesian_example.cpp as an exercise to better understand setting up an environment and use (Python) Tesseract. I should be able to make a contribution to this PR or the repo by adding this when most of the structure is there.