Prof. Dr. Robert Katzschmann

Research area

Real-World Deployed Soft Robotic Fish

Robert created SoFi, a fully integrated soft robot operating autonomously in the ocean to explore the behaviors of real fish (Katzschmann et al., Science Robotics, ‘18). SoFi’s actuated biomimetic soft tail mimics the propulsive undulation of real fish to unintrusively spy on marine life. Four publications on SoFi (Katzschmann et al. ‘16/‘16/‘18; Marchese et al. ‘15) have >1300 citations and led to invited talks at TechCrunch Sessions ’17, RoboSoft ‘19, SciFoo ‘22, NCCR Bioinspired Materials ’22, and TED ’22. The work was covered in the NY Times, Wall Street Journal, National Geographic, BBC, etc. 

Robert proposed the computational co-optimization of SoFi’s design and control. His team developed a fast differentiable method for co-designing a soft robot's geometry and controller for multi-objective problems (Ma et al., SIGGRAPH, ‘21). The differentiable approach allows researchers to automatically optimize the material and control parameters of their soft robotic designs (Zhang et al. IROS ’22, Gravert et al. IROS ‘22). Coupling deep-learning fluid models with continuum solid mechanics quickly optimizes the controls of soft robotic fish in multi-physics simulations (Nava et al. ICML ‘22). The work led to an invited talk at RoboSoft ’22.

Proprioceptive Robotic Hands

Robert creates dynamically controllable soft robotic manipulators that despite their large degrees of freedom can reliably perform object grasping (>160 citations, 1 invited talk) (Katzschmann et al., Soft Robotics, ‘15). He co-developed the Printable Hydraulics method that creates functional hydraulic soft robotic grippers in single prints (>210 citations) (MacCurdy et al., ICRA, ‘16). He designed and built a casted soft hand with integrated sensors, that could for the first time grasp and identify objects based only on its internal measurements (>300 citations) (Homberg et al, Autonomous Robots, ‘18). He then demonstrated the first proprioceptive object manipulation with a fully printed soft robotic hand (Best Paper Award, >50 citations) (Truby et al., RoboSoft, ‘19). His work on soft robotic hands led to five publications cited >825 times and invited talks at IROS ’15, ICRA ’16, GRC Robotics ’20, and ICRA ’21. In ’15, the BBC, Scientific American, Popular Science, etc. wrote about this work. At Amazon, he created soft suction grippers (“Robin”) that have already manipulated millions of packages (patent US10913165B1, ‘19).

Dynamic Control of Soft Robots

Robert created soft robotic arms and dynamic control algorithms that allowed for fast motion despite infinite degrees of freedom (Katzschmann et al., Robosoft, ‘19). He conceived an impedance and dynamics model that enabled the application of control techniques developed for rigid robots to the dynamic control of soft robots (Della Santina et al., IJRR, ‘20). This led to >500 citations in four publications, invited talks at RoboSoft ’18/’19, a workshop at IROS ‘18, and a special issue at IJRR.

He conceived the first proprioceptive soft arm that estimates its state using capacitive flex sensors and a fast dynamical model; soft robots can now move in visually occluded environments that disallow exteroceptive measurements (Toshimitsu et al., IROS, ‘21). He then conceived a dynamic controller that allows soft robotic arms to perform a variety of real-world tasks such as picking, throwing objects, drawing, or avoiding obstacles (Fischer et al., AIS, ‘22). Within two years, his group published six times on this topic and co-organized a RoboSoft ’22 workshop. As a result, Robert was invited for talks (Hamlyn Symposium ’21, NCCR Bio-Inspired Materials ’22), keynotes (ICCAR ’22 and ICAARS ’22), and advanced Ph.D. schools (SMART Winter School ’21, Deformation in Robotics ’22).