This paper describes an evolutional computation that gives an autonomous mobile robot skill the ability to avoid obstacles and search for the shortest route in a path planning problem. We designed a robot simulator in order to provide it on a real robot, with a CCD camera giving a broader view by setting it up on the top of the pole put on the real robot. Therefore, the simulator robot can measure approximately 5 directional distances from obstacles. To make the robot receive inputs (V_0, 〜V_4) from 5 directional image's sensors, the robot's functional nodes were used in non-terminal sets and it's sensor nodes were used in terminal sets. Particularly, we improved functional node Movewj (V_0, 〜V_4) that can move back after switching the steering angle to the inverse direction. The robot, implemented with the new functional node, has reached a goal on a real map using genetic programming.