Hello I am Mohammad Tahboub, I am developing a project to resolve an hypercube ( this problem is similar to resolve a cube of Rubik ), for example from a state A with one move i can reach a state B for the same object.

I need to find a way to teach to one AI to find and return the leatest numbers of moves to complete the task. I thought to use LSTM for NLP ( bidirectional ), but to be honest I have no idea if it is right.

I think I could do that simple using a brute-force algorithm adding it a database but in the same time I’d like to prove myself i learned well from coursera about AI, do you have any idea about that?

I studied all states and i know all new states for all moves, I want only to get the last numbers of correct moves to reach a state I want, because I know it’s possible.

do you have some courses about this type of Reinforcement learning?

A with move 1 get B with move 2 get C

A with move 2 get C with move 3 get D

A with move 3 get D with move 3 get F

A to F is ( 3, 3 ) ← stupid example sorry

I need Reinforcement learning also to resolve the problem to move in one open world game ( made for example by a sandbox game = like TextWorld world for make an example ).

Only to find a right hypothetical sequences.

How to implement it will be my problem, for now i need a mind stronger than mine ( possible without brute-force ).

Best regards

Samir