PID-LRLES controller boosts soft robot swimming accuracy in dynamic flows

A team of researchers from multiple institutions has introduced a novel control method for soft robotic swimming that significantly improves tracking accuracy under varying flow conditions. The approach augments conventional PID control with a Linear Repetitive Learning Estimation Scheme (PID-LRLES), which generalizes integral action from constant to periodic references while avoiding long-term instability. The controller was tested on a soft robotic swimmer in a recirculating flow tank at five bulk flow speeds ranging from 0 to 32.6 cm/s. Closed-loop experiments used an embedded soft capacitive bending sensor operating at 1 kHz, with controller gains tuned once in static water and held fixed across all conditions. The PID-LRLES tracked the periodic bending-envelope reference more closely than the PID baseline and significantly reduced inter-trial variability, with a paired Wilcoxon signed-rank test p-value of 1.8 × 10⁻⁴ (n=25).

The key insight is that embedded soft proprioception (sensing body deformation) and cycle-to-cycle learning work together: the sensor exposes periodic hydrodynamic bias, while the learning term absorbs it over recent oscillation cycles. This combination reduces flow-dependent control-induced variability, providing an enabling layer for future robophysical studies that aim to isolate the effects of morphology, sensing, and environmental flow on aquatic locomotion. By enabling more consistent and precise control, the PID-LRLES approach opens the door for embodied intelligence research in soft robotics, potentially leading to more robust autonomous underwater vehicles and bio-inspired robotic platforms.