Rolling-diaphragm hydrostatic transmission for the remote actuation of high-performance robots

The capability of safely operating in crowded, unstructured, anthropic environments led to a rapid spread of force-controllable collaborative robotic arms. However, the conflicting requirements posed by the traditional industrial applications in terms of operating speed and desired impedance still represent an unsolved trade-off. Rolling Diaphragm Hydrostatic Transmissions (RDHTs) are an ideal technology to merge collaborative and industrial robots in a unified advanced technological paradigm. Shaping a suitable open-loop response, regardless the applied control, by embedding requirements of safety and performance in smart mechanical structures is the key principle that drove this study: lightweight design and backdrivability are considered as the ultimate approach to the design of the next generation of human-like robotic arms. RDHTs are simple, passive systems that reflect torques across long distances in the form of fluid pressure and enable the remote positioning of electric motors, whose mass is taken away from the robot. Remotely located low-inertia direct or quasi-direct drive actuators, transparent and stiff torque transmission of the motors’ action to the joints, and lightweight links are the main ingredients to create safe robots with low mechanical impedance, high specific power, excellent backdrivability and large force bandwidth. In this thesis, a modular robotic joint based on novel hydraulic cylinders is developed; the cylinders implement rolling diaphragms and feature a minimally constrained floating-bonnet architecture that enables zero-leakage, low-friction operation. The joint generates a maximum rated torque of 25 Nm and exhibits a static friction value of just 0.24 Nm (0.96% of the maximum rated torque). Moreover frictional properties are independent of the applied load, which is a favorable feature inaccessible to the widely-used cable-based transmission systems. Exploiting a low-cost pressure sensing technique, a Smith-predictor-based joint torque control is developed to achieve enhanced torque setpoint regulation and high-quality physical Human-Robot Interaction (pHMI). Experimental tests show a reduction in backdriving torque of 67% and in settling time of 95% with respect to the open loop. The open-loop response is still largely acceptable over the whole range of frequencies that characterize realistic scenarios of manipulation and interaction. An advanced configuration of the robotic joint equipped with pressure sensors is then integrated in a remotely-actuated planar robotic arm. Excellent force controllability is confirmed in collaborative manipulation tasks that involve the displacement of a heavy payload over its entire workspace. The novel layout of the rolling diaphragm cylinders is further challenged by developing a miniaturized pneumatic version of the transmission system with intrinsic series-elastic properties for the remote actuation of the ankle joint of an agile hopping robotic leg. Finally, an effective solution for implementing programmable physical damping in hydrostatic transmissions is conceived.