Conception d'une Interface de Pilotage d'un Cobot

Abstract

In the present industrial context, there are still many tasks that cannot be fully automated, and that require human workers to do painful tasks. Robotic assistance may be a good solution to reduce the operator's force intensity needed to accomplish such tasks, and thus avoiding musculo-skeletal disorder cases. If we allow an operator to cooperate directly with a robot handling a tool (or an object), then it will be possible to consider helpful assistance functions such as weight compensation, force augmentation or virtual guidance. Industrial robots may be relevant candidates for such a cooperation because their mechanical capacities are very complementary to those of humans : they are especially optimized to supply high forces, to be very stiff and to be very accurate. However, high friction in the joints and high inertia of these robots conflict with a smooth and light interaction with the operator. Therefore, using a force sensor at the robot's end effector in a force control loop is essential. One shows that stability of such a controlled system highly depends on the mechanical impedance opposed to the robot when the human operator handles its end effector. We base our study on a robot model taking into account a structural flexibility between the actuator and the force sensor, which is the source of stability limits. To optimize the resulting highly constrained performance / stability compromise, we suggest a control law that takes into account the impedance characteristics of the hand grip thanks to a specially designed interaction handle. To characterize the hand grip, it has been necessary to develop a sensor capable of mapping the pressure applied to the round surface of the handle. We describe the architecture and operation of such a sensor, and show that it can also be used to cover the body of the robot. We suggest a solution to prevent the mobile parts of the robot to threaten the safety of the user, by increasing their sensitivity to physical interactions with the operator or the environment. We finish our study by considering a tripartite interaction where the robot amplifies the effort that the operator applies to the handle. By adapting the theoretical tools commonly used in the field of teleoperation, we conduct a comprehensive study on the limits of stability and performance of this system. All these improvements are based on theoretical developments supported by experiments on a real industrial robot.