Soft-Body Snake Robots Slash Underwater Energy Use
Soft robotics, built from compliant materials, has emerged as a transformative field with applications ranging from industrial manipulation to medical rehabilitation. Among its most compelling prospects are biomimetic underwater robots, which emulate the swimming mechanics of aquatic animals. While such designs promise efficiency akin to nature’s own swimmers, quantitative studies comparing their energy use to rigid-bodied counterparts have been scarce.
A recent investigation addressed this gap by simulating two snake-like underwater robots—one soft-bodied, the other rigid—matched in mass, dimensions, motor capacity, and degrees of freedom. Both employed anguilliform swimming, a mode characterized by undulating motion along a slender body. The soft robot’s segments incorporated dielectric elastomer actuators (DEAs), modeled in MuJoCo to capture their deformable behavior. Each of its five segments contained multiple active and passive joints, yet maintained a single degree of freedom per segment through synchronized actuation. The rigid robot, with six segments and five joints, was configured to match the total actuation range and torque capacity of the soft design.
The simulation environment replicated water density at 1,000 kg/m³ and viscosity at 0.0009 Pa·s, with control and simulation frequencies set at 50 Hz and 100 Hz respectively. Two control strategies were deployed: a gait equation controller using grid search to sweep through torque amplitude, spatial frequency, and temporal frequency parameters; and a deep reinforcement learning (DRL) controller based on proximal policy optimization (PPO). The DRL approach allowed exploration beyond predefined motion equations, guided by reward functions tuned either for minimal energy use (R1) or for achieving a target velocity (R2).
Grid search produced 3,600 distinct gaits per robot. Scatter plot analysis of average velocity versus average power revealed a pronounced efficiency advantage for the soft-bodied model. At an average velocity of 0.024 m/s, the soft robot required just 19.60% of the power consumed by the rigid counterpart—a reduction of 80.4%. This advantage widened at higher speeds. DRL training converged within 500 iterations, with results showing that soft robots more readily learned energy-efficient gaits, while rigid robots struggled to approach the efficiency boundary defined by grid search.
Additional experiments varied the rigid robot’s joint rotation range and torque limits to assess their impact. Rotation ranges outside the original ±36° degraded performance, and torque range adjustments yielded no better gaits than those already found via grid search. These findings indicated that the control space had been thoroughly explored, reinforcing that the efficiency gap stemmed from body compliance rather than control parameter bias.
Analysis of selected gaits with similar velocities and minimal power consumption provided further insight. Velocity profiles of the rigid robot exhibited greater fluctuations in center-of-mass speed, implying higher drag forces during swimming. In contrast, the soft robot’s continuous body deformation produced smoother velocity curves, reducing hydrodynamic drag. The compliant structure allowed efficient backward transport of water with minimal resistance, a trait consistent with observations in biological swimmers.
The study’s methodology combined precise mechanical modeling with advanced control optimization, demonstrating that soft-body dynamics confer substantial energy savings in underwater locomotion. By isolating variables—mass, size, torque capacity, and degrees of freedom—the researchers ensured that differences in performance arose from body compliance alone. The results align with prior theoretical work suggesting that fluid-inertial effects enhance propulsion in deformable bodies, but extend this understanding with rigorous quantitative comparison.
Such findings have implications beyond aquatic robotics. Energy-efficient motion through compliant structures could inform designs in aerial drones, terrestrial search-and-rescue robots, and planetary exploration rovers, where power constraints are critical. The demonstrated synergy between soft-body mechanics and modern control algorithms like DRL underscores a pathway for engineering machines that move not only with agility but with frugality in energy use.
