Mechanical and Civil Engineering Seminar

Despite robotics has made tremendous progress in perception and planning, robots are still far away from being as dexterous and sensible as animals or humans. I argue that a solution to this problem requires a departure from the classical paradigm of sensor-based planning on precise robots consisting of stiff joints, links, and gears to a new class of robots made of materials that tightly integrate sensing, actuation, computation and communication. Such a material provides the potential to off-load high-bandwidth sensing, processing, and control into the material, which can then react autonomously to its environment. I dub these novel composites "robotic materials". Robotic materials are enabled by recent advances in smart polymers, desktop manufacturing systems and miniaturization of computers. The challenges in creating robotic materials lie at the intersection of computer science, controls, and materials engineering and - from a CS perspective - require advances in distributed algorithms for signal processing, control and routing of information. I will illustrate these challenges and recent advances by my group using three systems that emphasize high-bandwidth sensing, distributed information processing, and distributed feedback control in robotics materials: (1) a soft robotic skin that can locate and classify textures by locally sampling, processing and classifying 3kHz vibration signals and route relevant information to a central processing unit using multi-hop networking. (2) A modular building block for creating intelligent walls and facade systems that can recognize complex gestures spanning multiple building blocks. And (3) variable stiffness composites that can assume arbitrary shapes using simple actuation and local feedback control.