Tactile Grasping for Domestic Service Robots : Simulations, Experiments and Hand Design

Abstract: This thesis presents methods and tools for robotic grasping. The application in mind is service robots for a home environment, although the work also has applications in other areas. Service robots must be capable of manipulation, i.e. able to move objects from one place to another, in order to be truly useful.There are many important technologies that all need to function together to enable robotic grasping. The robot must be mobile, have a robot arm, be able to navigate, and be equipped with vision and/or other sensor modalities. To be able to exploit all these features, the robot must be able to use its skills in an intelligent and safe manner even if its surroundings change over time. It must be able to create a representation of the world, reason about what to do, and be able to control its body so that the desired task can be performed. The focus of this thesis is the end-effector — the robot hand — and its control.An intuitive formulation of a hybrid force/position grasp controller is proposed. The method is applied to three different robot hands in simulation and to one robot hand for real life experiments. Dynamic simulation using a modified version of the grasp simulator Grasp-It! is shown to be useful in the development of control algorithms and robot hardware. The simulator is also used to establish the required accuracy for object position that ensures that an adequate grasp is reached. It is also shown that simulation in combination with heuristics (rules) can be used to choose the most suitable grasp among certain candidate grasps.The design of an under-actuated low-cost robot hand capable of hybrid force/control grasping is presented and its performance evaluated.The results are discussed and related to the state of the art, the grasping problem, and common issues within mobile manipulation. The conclusion is that for several manipulation tasks that may occur in a home environment, an intuitive control formulation and low-cost hardware can offer secure force-controlled grasping. Using simulation of hybrid force/position control, it is possible to establish task and system requirements that ensure that an adequate grasp is achieved.