Résumé
In this paper, we address the problem of long travel mobile manipulation for steerable wheeled mobile robots (SWMR) operating in human shared environment. On one hand, a small footprint is required while maintaining a fixed arm configuration, to make robot motion predictable for near individuals during the long traverse. On the other hand, redundancy resolution poses a challenge since there is no direct kinematic mapping between the task and joint spaces for SWMR. Hence, we propose a redundancy resolution algorithm that enables switching between 3 modes of operation based on the Euclidean norm of the motion task error. In particular, we employ a floating base model for the mobile platform, and enhance the end effector motion performance by predicting the error between such model and the actual (SWMR) one. Such error is then compensated using the highly responsive arm manipulator. The proposed methodology is successfully validated in simulations on a Neobotix-MPO700 SWMR with a Kuka LWR-IV manipulator mounted on it.