In Robotics, one of the core problems is motion plan ning towards a specified goal. Today, there are different solutions; however, most of them require complex models of the environment, which are not intuitive at all, and force the programmer to under stand the operating mode in detail of the robot she /he is working with. Furthermore, most algorithms use i naccurate information which cannot be relied upon. This paper presents a robot motion imitation approa ch, using Bayes networks and machine learning, which does not require much knowledge, neither an e nvironment model. The proposed method defines a Bayes network based on the distance detections near the robot. A learning phase provides the robot with a path for solving a motion planning problem and fo r processing the distance detections and velocities in order to learn how to control the robot. Finally, t he robot can reproduce the learned path by using a probabilistic inference algorithm. The proposed met hod was tested both in simulated and experimental environments with a differential mobile robot.