In this paper we propose a cognitive robotic system that utilizes computational psychology (the Soar cognitive architecture) and an obstacle avoidance method (modified dynamic window approach) in ROS (Robot Operating System) platform for controlling a mobile robot. This system is applied to perform a task of human-following, aiming to help the robot navigate itself to the target person avoiding collision. A cognitive agent based on Soar cognitive architecture is created to reason its current situation and make decisions on movement direction such as go-straight, turn-left or turn-right, whereas the dynamic window approach is modified to avoid collision by computing appropriate velocities for driving the robot motors. To the end, a part of implementation is presented to describes how the system works.