Current Progress on Control Strategies for Underwater Soft Robots: A Comprehensive Review

Abstract

Underwater soft robots (USRs) operate in highly dynamic, unstructured, and sensor-limited environments where fluid-structure interaction, compliance, and nonlinear actuation significantly complicate closed-loop control compared with both rigid underwater robots and terrestrial soft robots. As a result, classical underwater control techniques and conventional soft robot control approaches require substantial adaptation to address challenges such as uncertain hydrodynamics, limited onboard sensing, energy constraints, and slow actuation dynamics. This review provides a focused and critical synthesis of recent control strategies developed specifically for USRs. Literature published between 2019 and 2026 (till April) was systematically screened and analyzed, and dominant approaches were categorized into data-driven methods (reinforcement learning and neural networks), model-based methods (sliding mode control, model predictive control and proportional-integral-derivative), and hybrid strategies. For each category we compare representative controllers in terms of actuator modality, control objectives and operational environment. Key limitations are highlighted including model inaccuracy, sensing constraints and scalability, followed by prioritized research directions such as controller actuator co design, data efficient learning and robust control under severe environmental uncertainty. By offering a decision-oriented perspective rather than a cumulative survey, this review aims to guide the selection and development of control strategies tailored to the unique requirements of USRs.