Soft robots have the potential to address fundamental challenges in autonomy when interacting with unknown or uncertain environments by leveraging passive compliance and material intelligence during contact. This challenges the traditional “sense-think-act” feedback loop while maintaining some sense of autonomous decision-making. Yet the alternatives to traditional feedback control are unclear. Should dynamical systems analysis be retained, but with different objectives? Should we forgo traditional control, incorporate high-level perception and AI-driven world models only, and rely on material intelligence for low-level motions? Do soft robots need computation during their decision-making, at all?

To date, most successful work in control for soft robots has focused on adapting methods such as motion planning and trajectory tracking, which may not align with the intended physical intelligence of unplanned motions. This workshop convenes researchers in dynamical systems, AI, and mechanical design to review and compare possible framings of the question: What is the role for feedback control in soft robot autonomy, if there is one? No consensus currently exists about the value of precision and low-error motions in soft robotics, nor what other mathematical formulations of decision-making might truly work with softness rather than fight against it. Many possibilities exist for a role for control – from high-level logic, to learning and imitation, verifiable safety, and combinations therein – yet no prior venue has gathered experts to compare and formalize the alternatives in this landscape. And with the advent of generative AI, it is unclear if a lack of formalization in soft robot control is a limiting factor in autonomy, or if there is a more fundamental issue in defining “autonomy” and applying engineering principles to it.

Through invited talks, panel breakouts, poster sessions, group networking sessions, and interactive feedback from attendees, this workshop seeks to set a vision for autonomy and control in soft robotics for the next decade.