To coexist with human, a robot has to avoid obstacles based on human-like flexible decision-making. In this article, we ran the robot manually in outdoor “Tsukuba challenge” course, took time-sequential images with the camera, and recorded the point cloud with LiDAR. By associating images with point clouds, we calculated the direction and velocity of obstacles on the screen. We recorded the angle and speed when a human operates a robot while looking at the screen to avoid the moving obstacle on a developed simulator. Using those, fuzzy rules to decide the moving direction and control speed at every moment to avoid obstacles were derived as follows: as the input variables, distance to obstacle (x₁), angle to obstacle (x₂), speed of obstacles (x₃), and the direction of movement of obstacle (x₄), are adopted. As the output variable, steering angle (y₁) and control speed (y₂) of robot is adopted. Based on fuzzy-neural networks method, two network of 4 inputs (x₁～x₄) and 1 output (y₁ or y₂) is prepared. Fuzzy rules are obtained so as that the network reproduces the trajectories of simulation experiment with minimum errors. The obtained fuzzy rules successfully produce controls which is similar to human ones.