Motion-planning problems are discussed for a mobile robot with an incomplete map. For the given start and goal positions, the robot will preplan a collision-free path based on the given map, this is known as the find-path problem. A solution of the find-path problem is presented for randomly shaped obstacles using an improved retraction method, namely, retraction with local replacement by visibility graph. Three different versions of a motion replanning algorithm are presented that patch the world model and regenerate a path from the current location to the goal location. In this scheme, the recomputation of a completely updated world model is avoided by patching the locally updated environment.>