In order for humanoid robots to work actively around humans, they are expected to move autonomously in public spaces. For achieving such mobile intelligence, so many researches have been conducted so far. However, most of them deal with only elements for constructing necessary mobile intelligence. While moving around public spaces, composite recognition including self-motion recognition, moving objects' motion recognition, and global localization is indispensable in a dynamic environment. In this paper, “Mikoshi” is adopted as a challenging task of coordinated navigation problem in a dynamic environment. “Mikoshi” is a Japanese traditional mobile shrine which is carried by multiple people cooperatively. In some festival, not only a shrine, but also a person is carried. For robots to imitate ‘Mikoshi’ in a public space, they have to solve not only cooperative locomotion problem, but also navigation in a dynamic environment. The purpose of this paper is to achieve such coordinated navigation problem by multiple humanoid robots. For solving this problem, an integrated system based on real-time 3D space recognition and coordinated motion generation algorithm for stable motion is developed. Also, we performed a verification experiment with path planning navigation of ‘Mikoshi’ in a dynamic environment.