In disaster scenarios, mobile robots that have the capability to explore in dangerous environments can offer significant assistance in search and rescue missions instead of rescue crews. The realization of such robot systems is depending on many technologies. One important future technology is autonomous navigation to replace the currently used teleoperation technology. To develop a 3-D path planning method, we focused that how to evaluate environment by skilled operators of rescue robots. From the observations of maneuvers made by skilled operators, it was found that attitude maneuvers that are turning/rolling maneuvers, are minimized on rough terrain. Therefore, in this research, we propose a path-planning method for autonomous mobile robot considering the instability of attitude maneuvers on rough terrain. The proposed method consists of three steps: "gradient calculation," "nodes detection," and "path evaluation." In the gradient calculation step, the gradients of terrain are calculated. Next, in the nodes detection step, the terrain is divided into flat regions and rough regions, and the center positions of each flat regions are identified; this is called the "node." Finally, in the path evaluation step, the best path from current position to the target destination via nodes is planned using the proposed evaluation method. The proposed method has been implemented in our simulator, and some simulation experiments were conducted to demonstrate a validity of our method.