I'm writing an algorithm which includes a Traveling salesman problem (TSP) and a maze solving problem. Basically there are points inside the maze and we need to find the most optimal path to all those points and eventually exit the maze.
We started using an ACO algorithm to find the exit of the maze which works fine. But how would one integrate the TSP into it.
Our first guess would be reinforcement learning. Any ideas?
We figured out a way to do it. We decided to use a genetic algorithm where we encoded the order of each point in a chromosome. During each generation we ran the ACO algorithm for each chromosome and looked for the smalles amount of steps taken to reach the end goal.
It eventually converged or an iteration limit was hit.