search

dasvision0212 / 3D-Bin-Packing-Problem-with-BRKGA

An implementation ofr Biased Random Key Genetic Algorithmn for 3D Bin Packing Problem.
MIT License
56 stars 21 forks source link

about orientation of box for packing #2

Closed lujingqiao closed 3 years ago

lujingqiao commented 3 years ago

i'am sorry to ask you for one question: DFTRC-2 already get the best box orientation by maximize DFTRC, why later need to again re-cal orientation using VBO(first alternative)? re-cal box orientation can damage the calculated best box orientation? isn't it? thx in advance!

dasvision0212 commented 3 years ago

@lujingqiao You're right on this point. DFTRC-2 found an orientation that maximize DFTR, and using VBO seems like ignoring the effort.

Actually, the paper implemented both options: using DFTRC-2 orientation directly, or using VBO to re-cal orientation. It turns out using VBO to generate random orientation is empirically better. Though the paper didn't elucidate on this result.

My explanation is that by generating random orientation, you add more randomness (noise) to the decoding procedure, hence bigger solution space (more chance to find better packing sequence). In a sense, the orientation that maximize DFTR isn't always better. By allowing random orientation, you may find a better solution.

lujingqiao commented 3 years ago

thank you for reply