This article presents the development of a robotic ankle exoskeleton for human walking assistance. First, the hardware design of the robotic ankle exoskeleton is presented, including the mechanical structure, actuator design, and electronic system. A clutch-based nonlinear parallel elastic actuator (CNPEA) is designed to mimic the behavior of the human ankle joint and provide nonlinear stiffness output. In addition, an adaptive iterative learning control algorithm based on hierarchical reinforcement learning (HRILC) is proposed to adaptively generate and track the desired ankle joint assistive torque. Finally, experiments are conducted to validate the effectiveness of the developed robotic ankle exoskeleton and the HRILC algorithm. Experimental results demonstrate that with the proposed HRILC algorithm, the root-mean-square error between the desired and actual ankle joint assistive torques is 0.35 Nm, with concurrent reductions in gastrocnemius and soleus muscle activities of 26.49% and 23.71%, respectively. Furthermore, the design of the CNPEA reduces the peak motor current by 40.45%.
Magnetically levitated planar actuators (MLPAs) offer contactless motion, high precision, and kinematic flexibility, making them promising positioning platforms for industrial manufacturing applications such as laser cutting, visual inspection, additive manufacturing, and painting; however, their integration in cooperative tasks with robot manipulators remains largely unexplored. This study proposes a cooperative framework in which a disc-magnet mover and a 4-DOF robotic arm execute a configuration-sampled, time-optimal planned path to complete a defined trajectory with orientation constraints. The proposed path planning algorithm utilizes a geometrically modeled arm inverse kinematics together with a filtered discrete arm workspace to generate a global search graph, whose states are represented by locally feasible configurations found via a Jacobian-based redundancy resolution. To validate the flexibility of the MLPA against conventional implementations, a trajectory-aligned mover baseline is developed, in which the mover is constrained to remain fixed during the interaction phase with the manipulator. Experimental validation on three types of trajectories with a real MLPA system showed promising results from the proposed cooperative path planner compared to the baseline, with a faster completion time of 52.82±0.37s against the baseline of 64.7±0.26s. The cooperative path planner states were successfully executed, obtaining a maximum positional tracking RMSE of 0.54 mm with a standard deviation of 0.43 mm and an attack angle RMSE of 0.18°, with a standard deviation of 0.09° for a line trajectory. The second trajectory confirmed the need for cooperative mover-arm operation to obtain a feasible interaction solution. The final sinusoidal trajectory showed the high dynamic tilting range achieved from the mover to enable the lower DOF arm to complete the proposed path, demonstrating the potential of MLPA-robotic arm integration for flexible manufacturing on kinematically limited industrial lines.
BIFLEX is a flexible underactuated gripper inspired by the human thumb–index pair and designed for distal precision pinch and proximal adaptive holding. Building on a previously developed soft-hybrid Fin-Ray finger, this study combines finger-geometry optimization, four-bar kinematic synthesis, and shared single-motor actuation through a four-bar and tendon–pulley transmission. Parametric analyses guide rib orientation and compliant-joint geometry, while coupled finite-element analysis quantifies tendon forces, joint stresses, and crank torque during closure. The simulated maximum tendon force was approximately 2.9N, and maximum compliant-joint stress remained below approximately 3.7 MPa. Grasping experiments covered 23 object types, with ten trials per object and success defined as retention for at least 10s without visible slip. Additional tests evaluated three cylindrical objects at motor commands of 120°, 150°, and 180°. Separate load-holding experiments demonstrated stable loads of 6.5 kg in an inclined single-finger configuration and 8 kg in a vertically inverted two-finger configuration. Cyclic tests without object contact characterized angular hysteresis, repeatability, and out-of-plane motion. An SLS-type angular model described the measured creep response, with fitted amplitudes of 0.336°–0.742°. These results characterize the capabilities and configuration-dependent limitations of a single-motor gripper targeting two taxonomy-inspired grasping modes.
This paper explores the practical benefits of an advanced numerical differentiation method through two applications: the identification of a robot’s numerical parameters and the control of its motors. The experimental platform under investigation is a cable-driven parallel robot (CDPR). The paper begins by presenting preliminary work to identify several dynamic parameters of the CDPR, Cable Robot for Agile operations in manufacturing FaciliTies (CRAFT), using a specific orbital trajectory of the platform. This orbital trajectory is designed for the platform to achieve pure translational motion in space. Thus, as the platform does not rotate around its own axes, this simplifies the dynamic model of the platform by canceling out the terms for angular velocities and angular accelerations of the platform. This trajectory lies in a plane and enables the identification of four dynamic parameters: the platform’s mass and the first three moments about the three principal axes. From the experimental data, the linear velocity and acceleration of the platform, which are crucial for identifying the CDPR’s parameters, are computed using semi-implicit discrete-time homogeneous differentiators, for which a detailed convergence proof is given and compared with the conventional finite difference method. The identification results show that, despite the complexity of CRAFT, it is possible to identify its key dynamic parameters using both differentiation methods. Furthermore, in the long run, online identification of CRAFT during handling tasks is envisioned. In the second phase, both differentiation methods are employed and compared for closed-loop control of the eight motors, ensuring that the platform tracks an orbital trajectory at three different movement velocities.
X-by-wire mechatronic actuators are gradually replacing traditional chassis technologies in the automotive field. Following this trend, fully active electromechanical shock absorbers can provide controllable forces within the vehicle suspension in the four quadrants of their force–speed characteristic. Although linear electromagnetic actuators could seem adequate to perform this task, their limited force density makes them unsuitable in real suspension systems. Thus, rotary electric machines combined with devices able to transform the motion from rotary to linear are required. When combined with a lever mechanism, the linear motion of the suspension can be transformed into a limited angular rotation; the electric machine needs a speed reducer to increase rotor angular motion and reduce its torque demand. This functionality can be attained by means of a gearbox, which plays a crucial role in the actuator performance. Among the possible alternatives, cycloidal drives can reach high transmission ratios in compact sizes, while also guaranteeing low backlash and noise. They represent a promising solution when compared to other technologies, particularly in terms of noise, vibration and harshness. This paper presents the application of a cycloidal drive designed for active shock absorbers. A prototype has been studied, designed, integrated with a rotary electric machine and tested experimentally. The device is compared to a planetary reducer, which represents a benchmark solution previously adopted for the same application. Experimental results demonstrate that, while the efficiency of the cycloidal drive is lower in comparison to the benchmark solution, significant benefits can be obtained in terms of noise and vibration. This outcome is of particular relevance in the context of chassis actuators, where acoustic emissions are critical to the user experience.
