Problem
Research questions and friction points this paper is trying to address.
sim-to-real transfer
robotic fish control
low-fidelity simulation
soft robotics
zero-shot learning
All You Need Is Low Fidelity: Zero-Shot Sim-to-Real of Learned Robotic Fish Control
🤖 AI Summary
This study addresses the inherent trade-off between high-fidelity simulation costs and zero-shot transfer in soft robotic fish control by proposing an innovative paradigm that embeds physical constraints within the controller rather than the simulator. Methodologically, it employs a low-fidelity, stateless quasi-steady fluid model to reduce computational overhead, integrates staged system identification with reinforcement learning to train closed-loop policies, and utilizes a band-limited rhythmic trajectory generator to ensure the physical feasibility of output actions. Experimental results demonstrate that moderately reducing simulation fidelity paradoxically enhances generalization. A single policy achieves goal reaching, disturbance rejection, and out-of-distribution trajectory tracking in an outdoor pool without parameter tuning, successfully realizing zero-shot sim-to-real hardware transfer.
📝 Abstract
Complex tasks for underwater robots remain limited by the capabilities of their controllers. Learning a better one for a soft, underactuated robotic fish trades simulator cost against fidelity. We show that an intentionally low-fidelity simulator is enough: a stateless, quasi-steady fluid model with no wake and no added-mass history suffices to learn a \emph{general}, closed-loop controller that transfers to hardware without tuning. Our platform is a soft, single-motor, tendon-driven fish whose policy observes only what the hardware can measure. A staged pipeline grounds the simulator in two independent identifications, fixing the tail dynamics and a stateless fluid model; the policy then acts through a band-limited rhythmic trajectory generator rather than commanding the tail directly. Deployed unchanged in an outdoor pool, a single policy performs closed-loop target reaching, disturbance rejection, and out-of-distribution target acquisition and tracking. The transfer rests on the constraint rather than the fidelity: the generator cannot leave the band over which the fluid was identified. This raises the question of how much of the physics can reside in the controller rather than in the simulator.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Zero-Shot Sim-to-Real
Low-Fidelity Simulation
Soft Robotic Fish
Band-Limited Trajectory Generator
Closed-Loop Control
🔎 Similar Papers
No similar papers found.
