variable-stiffness actuation

Antagonistic soft actuators—such as pneumatic artificial muscles (PAMs), hydraulically amplified self-healing electrostatic actuators (HASELs), and dielectric elastomer actuators (DEAs)—face fundamental challenges in decoupling torque and stiffness control during dynamic contact. Method: This paper proposes a unified force model and a bias–coactivation coordinate control framework, enabling an analytically derived inverse-dynamics-compensated cascaded controller. Contribution/Results: The approach achieves millisecond-level torque–stiffness decoupling across multiple actuator types for the first time, effectively suppressing model uncertainties and external disturbances while emulating biological impedance regulation. Simulations demonstrate a 200× reduction in soft-surface contact stabilization time, an 81% decrease in hard-surface impact force, and 100% decoupling stability—substantially outperforming fixed-impedance strategies (22–54% improvement). This work establishes a new paradigm for real-time adaptive impedance control in soft robotics, enhancing safety and environmental adaptability in human–robot interaction.

This work addresses the challenge of jointly evaluating dexterity and compliance for modular soft–hard hybrid robotic fingers actuated by either hydraulic or pneumatic means. We propose the first unified modeling framework applicable across both actuation modalities. Leveraging Jacobian mapping, the framework integrates the incompressibility assumption (hydraulic) with a nonlinear pressure–volume relationship (pneumatic) to derive consistent manipulability ellipsoids and compliance matrices. This enables systematic analysis of the trade-off between dexterity and passive stiffness under structural–actuational coupling. Experimental validation on the DexCo (hydraulic) and Edgy-2 (pneumatic) platforms demonstrates the framework’s ability to quantitatively capture significant differences in manipulability and stiffness between the two actuation schemes. The approach provides both theoretical foundations and quantitative tools for task-driven co-design of soft–hard finger architectures and actuators.

This work addresses the challenge of uncertain joint stiffness in flexible-joint robots, which arises from time-varying and aging characteristics of elastic components and significantly degrades model-based control performance. To overcome this issue, a novel adaptive control method is proposed that continuously estimates and updates the nonlinear torque–deflection relationship of each joint online. By integrating an implicit control law with an input-dependent regressor matrix, the approach transcends the conventional limitations of adaptive control frameworks designed for non-elastic systems. The method substantially enhances robustness and tracking accuracy for position-controlled flexible-joint robots under varying stiffness and motor positioning errors. Experimental validation on a platform exhibiting nonlinear stiffness characteristics confirms the efficacy of the proposed strategy.

Traditional variable-stiffness actuators suffer from bulky structures and excessive mass, limiting their applicability in unstructured scenarios such as prosthetics and humanoid robotics. To address this, this paper proposes a compact, three-degree-of-freedom parallel variable-stiffness wrist joint. Its novel redundant elastic parallel architecture employs four motor-driven actuators, enabling decoupled, independent control of position and stiffness while achieving significant miniaturization and weight reduction. Leveraging elastic actuator modeling, parallel kinematic analysis, and nonlinear dynamic system design, we develop a high-precision stiffness-adaptive controller. Simulation results demonstrate: (i) high positioning accuracy and strong disturbance rejection in stiff mode; and (ii) over 60% reduction in interaction force in compliant mode, markedly enhancing human–robot interaction safety and task adaptability.

Existing artificial muscles are predominantly unidirectional—capable of either contraction or extension only—hindering the realization of fully functional, biologically inspired antagonistic musculoskeletal systems. To address this, we propose a biomimetic antagonistic architecture that synergistically integrates non-extensible hydraulic amplification of static electric fields (HASEL) actuators with electrostatic clutches, enabling the first demonstration of bidirectional, seamless actuation and high-frequency synchronous operation (up to 3.2 Hz). The system incorporates an antagonistic tendon-driven framework coupled with a real-time synchronized control strategy, achieving full-range motion without positional drift or displacement loss. Furthermore, it exhibits cross-platform generalizability, readily accommodating other non-extensible artificial muscles such as McKibben actuators. This work significantly advances the functional completeness, dynamic performance, and biological fidelity of artificial skeletal muscle systems, establishing a novel paradigm for bioinspired robotics and rehabilitation devices.

This study addresses the challenge of rapidly customizing stiffness in compliant mechanisms due to geometric constraints and stiffness coupling. A Lego-like stackable planar compliant module design method is proposed, establishing a unified model through a novel modular stiffness configuration framework to achieve stiffness decoupling and flexible reconfiguration. The optimization employs a genetic algorithm for module configuration search combined with sequential quadratic programming for parameter refinement. Experimental results demonstrate that simulated stiffness deviations remain below 6.5%. Furthermore, a developed compliant wrist prototype achieves approximately 15° angular compliance alongside prescribed stiffness characteristics during high-speed motion, validating the effectiveness of the proposed approach.

This study addresses the challenge of online task-space stiffness adjustment in supportive continuum robots, whose passive stiffness is fixed and constrained by closed-chain kinematics, limiting adaptability to varying payloads. To overcome this limitation, the work proposes an active stiffness control framework that, for the first time in such robots, enables virtual Cartesian spring stiffness modulation based on position error feedback. A geometrically varying strain model captures the closed-chain dynamics and projects them onto the constraint-consistent motion subspace. By integrating sliding-mode control with Lyapunov-based stability analysis, the approach simultaneously achieves precise end-effector positioning and strict enforcement of closed-chain constraints. Simulation and experimental results demonstrate that increasing the stiffness gain significantly reduces payload-induced end-effector deflection, thereby enhancing directional stiffness, disturbance rejection, and positioning accuracy.

本文提出了一种基于气动肌肉的软外骨骼控制策略，通过仅依赖运动学感知实现无需特定任务识别的动态辅助，有效减少了交互扭矩和肌肉激活。

本文针对软气动执行器的非线性和不确定性问题，提出基于非最小坐标离散弹性杆模型的实时动态模型控制框架，结合准静态前馈逆模型、任务空间PI控制器和动态观测器，实现高精度控制。

This study addresses the unclear morphological mechanisms underlying adaptive stiffness regulation in the human wrist. We developed an anatomically accurate biomimetic soft forearm incorporating independent carpal bones, integrating multi-degree-of-freedom actuation with stiffness ellipse analysis to systematically characterize joint stiffness under diverse muscle activation patterns. Our findings demonstrate that carpal morphology predominantly governs stiffness modulation and elucidate the critical role of proximal carpal coupling in determining stiffness along the low dart-thrower’s motion (DTM) direction, successfully replicating human wrist stiffness characteristics. This work reveals the biomechanical principles governing stiffness regulation and establishes a novel paradigm for designing anthropomorphic dexterous robotic wrists.