Train an agent to navigate in a grid with traps and walls, with no prior knowledge of the grid layout. The grid size i variable. The agent is trained using TD-learning, with the Monte Carlo method. The model is positively rewarded when it enter the finishing state, negatively when it enters a state with a trap, or if it navigates to a previously visited state.

• Variable learning rate.
• Variable exploration rate.
• Indicate from which game, it wins.