Safe Autonomy for Soft Robots
Merging the embodied intelligence of soft materials with computational intelligence has the potential for robust robotic interactions in unstructured environments. However, attempts at control for soft robots have mostly focused on positioning of soft bodies in space, recreating the problems that originally motivated a transition to softness. Are we asking the right questions, and working toward the most fruitful goals? In this talk, I will make the argument that safe autonomy can untangle many paradoxes surrounding softness and physical robot-environment contact by avoiding precision as an objective. Supervisory control systems can enable soft robots to perform “teach and repeat” tasks within their actuator limits, and can prevent mechanical fatigue and degradation. Fault detection methods can ensure that novel soft actuators are used safely during feedback. And, set invariance methods such as control barrier functions can enforce task-space constraints on applied forces, over time. We demonstrate that these methods for planning, control, and automation can be applied real-time, in hardware, for both soft robotic locomotion and manipulation. Along the way, this work contributes insights into modeling of soft bodies in complex loading conditions, and practical integration of force sensing into soft mechanisms. This perspective of safe autonomy may help bring soft robots out into real-world deployment.
Bio:
Andrew Sabelhaus is an Assistant Professor in the Department of Mechanical Engineering and Division of Systems Engineering at Boston University. He received a Ph.D. from the University of California, Berkeley in 2019, an M.S. from Berkeley in 2015, and a B.S. from the University of Maryland in 2012, all in Mechanical Engineering. From 2019-2021, he was a postdoctoral researcher at Carnegie Mellon University. He worked with NASA Ames Research Center as a Visiting Technologist from 2015-2019.
Prof. Sabelhaus leads the Soft Robotics Control Lab at BU, which builds autonomy into soft and flexible robots. His research incorporates motion planning and control techniques for these robots’ soft bodies and artificial muscles, focusing on safe operation during human contact and in unknown environments. He was awarded the NSF CAREER Award in 2024, an Intelligence Community Postdoctoral Research Fellowship in 2020, a NASA Space Technology Research Fellowship in 2015, and an NSF Graduate Research Fellowship in 2012.