This paper explores the application of Markov Decision Processes (MDPs) in Reinforcement Learning to find a navigation policy in three-dimensional (3D) spaces. MDPs can define a comprehensive state space and guide the selection of optimal navigation strategies. The proposed methodology included the definition of the modeled environment, where each point was represented by a viable state, and the identification of possible actions, including movements in different three-dimensional directions. Furthermore, a reward function was designed that penalizes longer trajectories and encourages efficient collision avoidance. The results showed that the models were able to find a policy that allows the robot to execute the shortest path, both in the absence and presence of obstacles.
