This paper mainly discusses the design and implementation of semi-automatically simultaneous localization and mapping (SLAM). SLAM is an important technique for home service robots to move in unknown environments and built the environment map. Different from traditional SLAM method which is operated by a remote controller, the semi-automatically SLAM is operated by human-robot interaction. The semi-automatically SLAM method is composed of SLAM, obstacle avoidance strategies and following technique. With this method, the robot can follow human operator walking around the surrounding, and build the environment map at the same time. The SLAM system is built using the Iterative Closest Point (ICP) algorithm. The ICP algorithm estimates the pose of the robot and compares the prior built map with current laser information to iteratively revise the environment map. Furthermore, Q-learning is applied for obstacle avoidance during navigation. After learning, the robot can navigates smoothly and avoid obstacles automatically. The proposed methods experimented in the laboratory and in the RoboCup Japan Open 2014 competition. The validity and efficiency of the semi-automatically SLAM for the home service robot are demonstrated.