MIP solver allows path splitting

[AlexisSchOrange]

Hello,
Thanks for the very nice tool, I will probably use it for my thesis.
I've been looking at your MIP implementation a bit, it seems that it allows path splitting (which it should not if I have understood the doc):

`
from virne.base import BasicScenario
from virne import Config, REGISTRY, Generator, update_simulation_setting

def run(config):
print(f"\n{'-' * 20} Start {'-' * 20}\n")
# Load solver info: environment and solver class
solver_info = REGISTRY.get(config.solver_name)
Env, Solver = solver_info['env'], solver_info['solver']
print(f'Use {config.solver_name} Solver (Type = {solver_info["type"]})...\n')

scenario = BasicScenario.from_config(Env, Solver, config)
scenario.run()

print(f"\n{'-' * 20}   Complete   {'-' * 20}\n")

if name == 'main':
config = Config(
solver_name='mip',
# p_net_setting_path='customized_p_net_setting_file_path',
# v_sim_setting_path='customized_v_sim_setting_file_path',
)
config.update({'verbose':1})
Generator.generate_dataset(config, p_net=True, v_nets=True, save=True)
run(config)
`

I think the paper from Chowdury et al. allows for path splitting. A formulation that can be found in some papers, with variable x and y (instead of f) allows to prevent path splitting (although it makes the problem much harder).

[GeminiLight]

Thank you for your feedback. I agree that implementing the MLP solver is challenging without allowing for path splitting. We should consider additional constraints on path routing and incorporate them into the solver's formulation. However, due to limited time and energy, these features may be considered for future development.

[AlexisSchOrange]

Thanks ! If I find some time I will try to write it and push it.

[GeminiLight]

Woohoo! Thank you so much! 😊