Contact sensing and mobility in rough and cluttered environments

Abstract

In this study, we explore mobility in a cluttered, uneven and dynamic environment where collisions are inevitable. In particular, we study a mobile robot's ability to precisely track a planned path on an inclined surface, then respond to unknown collisions by using force compliance and estimation of the obstacle's surface normal. In addition, once the contact disturbance disappears, the robot merges automatically back to its original planned path. This study aims at highlighting the need to safely deal with unexpected collisions with external objects and use them to move alongside their contour, opening opportunities perhaps to use contacts as supporting structures. At the same time, it aims to endow capabilities for precise maneuvering in the rough terrains. Our newly designed compliant omnidirectional robot is able to quickly respond to contacts through the use of current control and absolute position / orientation sensing. A whole-body controller developed by the Human Centered Robotics Lab is used to compensate for gravity and create new multicontact constraints in response to the collisions. To pursue the study, we built the mobile base and a multicontact experimental setup. The results show that our model-based controller successfully accomplishes the tasks with good precision while relying on data-driven contact estimation.