Hey!

I'm soon to start writing my masters thesis and I'm currently looking into how to solve a specific problem. I'm basically tasked with finding a solution to what could be considered floor planning for a warehouse. I'm not looking at shelves and single items within these shelves, but large objects which are restrained by a) other objects and b) certain other factors such as availability of certain in- and outputs (think gas, water, electricity etc) which are only available at certain areas in our warehouse. In short: Where is the best place to put an object (large objects, often bigger than 10x10 meters) in a warehouse (180x40m) with having restrictions in mind.

We have data on the positioning of other objects from the past 10 years (as in dimensions, positions and required mediums). Now I was wondering what the best approach would be to 1. analyze the data we have and 2. suggest possible locations for new objects in our warehouse.

Would this actually something ML would be useful for? I'm thinking that it sounds like something that would be possible using ML, as we have data to train our model and have relevant data available (think a abstract representation of the warehouse) when it comes to judging how good a suggested position actually is. Then again, floor planning and optimization has been done for quite some time before ML was even a thing.

I also figured that many relevant aspects and maybe correlations of objects placed in the past could easily be found using statistics, as all the data is available as a big spreadsheet. I don't want to say I solved my problem by using ML when in reality it is just glorified statistics.

My question boils down to: Is this kind of problem suited to be solved with ML or would I get better results by looking into "classic" floor planning? Thank you!