Description | Two approaches to make robots robustly recover from failure ABSTRACT: The essence of autonomy is the ability to persist in performing a desired task in the face of unexpected obstacles and internal failures. In the last few years we have worked on methods which allowed multi-legged robots to persistently move through the world despite lacking an a-priori model of themselves or their interaction with the environment. We rely on tools from geometric mechanics and differential geometry to present two such approaches, would share some common features. Both approaches rely on data-driven floquet analysis (DDFA) tools to create and maintain data-driven models of the interaction of the robot with the environment. With one approach, these tools are then used in conjunction with ideas from geometric mechanics to optimize a library of gaits that enable motion through the plane, and maintain the expressive power of that library even when the robot undergoes internal failures. With the other approach, the DDFA tools are used to derive virtual constraints that can be used in conjunction with both partial physical models and desired “Behavior Specifications” to create a framework that robustly recovers from many practically meaningful forms of failure. This talk will be of interest to roboticists, mechanical engineers, and physicists or applied mathematicians with an interest in mechanics. SPEAKER BIO: Shai Revzen is an Associate Professor in the departments of Electrical Engineering and Computer Science and of Ecology and Evolutionary Biology in the University of Michigan, Ann Arbor, and a core founding member of the Michigan Robotics Institute. He holds a PhD from the University of California, Berkeley in biomechanics, an MSc in computer science from the Hebrew University in Jerusalem, and did his post-doctoral work in robotics in the GRASP lab of the University of Pennsylvania. He heads the BIRDS (Biologically Inspired Robotics and Dynamical Systems) Lab, which focuses on the crucial role of mechanics in robot motion, and on the rapid manufacturing of robots and robotic mechanisms. The aim of BIRDS Lab is to extract core principles of biological robustness and control in mathematical form and recast them into robots. Shai's work has been published in engineering, biology and applied mathematics journals. Shai is also a co-founder of a company developing novel electro-cardiac diagnostics, is the author of several patents, and has extensive experience from over a decade in the tech industry in both Tel-Aviv and Silicon Valley. |
---|