Inefficient bin utilization with BinPackWithLeftovers in irregular

[SoftDev0208]

Hi Florian.
I checked your packing solver project, and it’s amazing.
However, I have an issue at the moment.
Your BinPackingWithLeftovers algorithm works very well for simple shapes, but it doesn’t perform well when dealing with many shapes or more complex ones.
This is input json.
packingsolver_input.json

Can you teach me if what's wrong?
this is the parameter.
.\packingsolver_irregular\packingsolver_irregular.exe --input packingsolver_input.json --certificate packingsolver_solution.json --time-limit 120 --objective BinPackingWithLeftovers --initial-maximum-approximation-ratio 0.03 --maximum-approximation-ratio-factor 0.8

[fontanf]

Hi,

The solver is currently not suited for instances with that many items in bins

[SoftDev0208]

The overall result is acceptable, but the optimization struggles with certain complex or special shapes.

[fontanf]

Please do not hesitate to share your other difficult inputs. That would be helpful to improve the solver

[SoftDev0208]

I think you should focus on the mix of shapes and arrangement sequence to maximize bin utilization.
The main goal is that he utilization rate of each bin must reach over 92% (excluding the last plate).

These are the current parameters.
"--time-limit", 300,
"--objective", "BinPackingWithLeftovers",
"--initial-maximum-approximation-ratio", 0.03 or 0.01),
"--maximum-approximation-ratio-factor", 0.8 or 0.75),

How can I set parameter for my goal?

[fontanf]

There is currently no option that would improve the solutions of the solver on this kind of instances

[SoftDev0208]

Got it. Then when can you update option for this kind of instances?

[fontanf]

I don't know, this is not simple to do. I have some ideas of what could be implemented to improve the performances of the solver on these instances, but it might not work at all. So, it could arrive in weeks, months, years or never

[SoftDev0208]

I think you should focus on the mix of shapes and arrangement sequence to maximize bin utilization as I mentioned over.

[SoftDev0208]

May I share other difficult inputs for you?

[fontanf]

Yes, sure. Ideally, if you can share a dozen of typical inputs of what you solve. That makes it easier to identify common structures to exploit in the algorithms

[SoftDev0208]

My inputs differ slightly from yours.
Anyway I'll share.

test.zip

[SoftDev0208]

I'll explain

reference.txt

important properties of my jsons.

[SoftDev0208]

hi how are you? Any good updates?

[fontanf]

Thank you for this data. Is it possible for you to convert them to packingsolver's format? I'm not sure I would convert them correctly