[vladimirralev]

But also "the memory" of the random walk can be encoded in a state itself. In your example you can just keep accumulating the visited nodes, so your state space will be now the space of tuples of nodes from your initial space.