This paper proposes an elliptical world-modeling mehtrod in order to secure the safety of the mobile robot from the unreliability of ultrasonic sensors and to estimate the lateral orientation of the detected objects. This method is advnatageous because the total amount of data that the robot utilizes is reduced when historical and new data are clusterd during each sensing period and also because the objects in the environment can be almost exactly recognized. This paper also proposes a new local path planning algorithm which is divided into four different situations and gudies the robot through the safe local path using dat acollected in elliptical form. This algorithm can help the AMR navigate by allowing it to sum up the virtual vector forces made form the ellipses in order to select a safe subgoal. This local path plannin gmehtod using the elliptical world model has been proved by several navigation experiments.