Multi-contact haptic exploration and grasping with tactile sensors

Haptic exploration has received a great deal of attention of late thanks to the variety of commercially available tactile sensors. While the majority of previous works consider control of a single contact point at a time, we tackle simultaneous control of multiple contact points on several links. In addition, we use information from the existing tactile signals to increase the number of points in contact. We demonstrate the usefulness of this form of control to speed up exploration, scanning and to compliantly grasp unknown objects. Our controller requires to know only the parts of the robot on which it is desirable to make contact and does not need a model of the environment besides the robot itself. We validate the algorithm in a set of experiments using a robotic hand covered with tactile sensors and arm. In a grasping application, the active adaptation of the fingers to the shape of the object ensures that the hand encloses the object with multiple contact points. We show that this improves the robustness of the grasp compared to simple enclosing strategies. When combined with an exploration strategy, our multi-contact approach offers an efficient use of tactile sensors on the whole surface of robotic fingers, and enables the robot to perform a rapid exploration of complex, non convex shapes while maintaining low contact forces. It is robust to variation in the approach angle and to changes in the geometry and orientation of the object. A control strategy for multi-contact exploration using tactile signals is proposed.It exploits all DOFs of high-DOFs robotic hands to make many contacts.Contact forces are kept low while creating new contacts and moving along objects.Fast exploration of complex shapes is performed in simulation.It can be applied to grasping unknown objects by enveloping them with the whole hand.