We propose odometry in cooperative multi-mobile robots by integrating conventional odometry and interrobot position sensor information. In our odometry, each robot is considered a moving landmark with imprecise location. Robots in the group locally estimate their own absolute positions based on conventional odometry and find the relative positions of each other using interrobot position sensors. They communicate and exchange information on local estimates and relative positions. The information is integrated decentralized based on the extended Kalman filter and robots improve their absolute positions. Simulation and experiments show that our odometry eliminates large robot location errors found in conventional odometry.