Docking is an essential yet challenging problem to the deployment of marsupial robotic system involving a mother robot and several child robots, as the technologies such as rail-line tracking and global vision measurement are not suitable for applications in unknown environments. This chapter presents a coordinated docking approach based on embedded vision of the child robot collaborated by the mother robot with an innovative lifting docking station. The child robot utilizes a complementary metal oxide semiconductor (CMOS) camera to capture and extract image features related to a benchmark attached on the docking station, and the internal correlation between its heading direction and the entry direction of docking station is then derived, which makes the docking process be concise and convenient. On this basis, a coordinated docking task model is built and the transition conditions among task states are presented. The precision analysis of the docking is also discussed. The experimental results prove the effectiveness of the proposed approach.