This paper presents an integrated control and transactional framework aimed at enhancing the stability, resilience, and automation of decentralized microgrid systems. The proposed approach combines an advanced Sliding Mode Control (SMC) scheme with blockchain-enabled smart contracts to address critical challenges associated with conventional control strategies, including slow convergence rates, susceptibility to disturbances, and limited scalability. The SMC is augmented with a nonlinear disturbance observer to provide fast transient response, robust disturbance rejection, and reduced control chattering under dynamic operating conditions. A closed-loop interaction between the SMC and blockchain layers enables continuous two-way communication. Real-time operational parameters, such as power imbalances, voltage deviations, and frequency fluctuations, are transmitted from the controller to the blockchain layer. In response, smart contracts autonomously trigger control adjustments and Demand Response (DR) actions, which are fed back to the SMC as reference inputs. This dynamic feedback loop enables the system to adapt to fluctuations and uncertainties in both energy generation and consumption, thereby ensuring consistent power quality and system stability. Experimental results confirm the system’s ability to maintain stability and improve power quality under varying operational conditions. Moreover, the system facilitates coordinated energy exchange among interconnected microgrids, thereby supporting the integration of local renewable energy sources and reducing dependence on centralized grid infrastructure.

Abstract Background Knee exoskeletons represent a significant advancement in wearable robotic technology, developed to enhance gait assistance and facilitate rehabilitation processes. These devices are increasingly used in clinical and research environments worldwide. Nonetheless, the existing empirical evidence supporting their biomechanical and functional outcomes lacks both breadth and clarity. Further investigation is essential to elucidate the specific effects of knee exoskeletons on gait dynamics and rehabilitation efficacy. Objective This study aims to systematically review the clinical and biomechanical evidence on knee exoskeletons used for gait assistance and rehabilitation. Methods A systematic search of databases—PubMed, Web of Science, IEEE Xplore, and Scopus—was conducted to identify eligible studies published between January 2015 and March 2025. Studies were included if they investigated knee exoskeleton systems in human participants for gait assistance and rehabilitation. Randomized controlled trials, observational studies, and experimental investigations were eligible. Exclusion criteria were multijoint exoskeletons, nonhuman studies, and simulation-based analyses. Two reviewers independently extracted study characteristics, participant demographics, device features, intervention protocols, and clinical and biomechanical outcomes. Risk of bias was assessed using RoB-2 (Cochrane Risk of Bias 2) for randomized controlled trials and ROBINS-I (Risk of Bias in Nonrandomized Studies of Interventions) for nonrandomized studies. Results Thirty-two studies met the inclusion criteria. Most investigations involved healthy adults in feasibility or pilot settings, while a smaller subset included individuals with stroke, spinal cord injury, or orthopedic conditions or pediatric movement disorders. Across studies, knee exoskeletons demonstrated short-term improvements in spatiotemporal gait parameters, including increased gait speed and step length, improved knee joint kinematics, and reductions in muscular effort or metabolic demand. However, substantial heterogeneity in device design, outcome metrics, and intervention duration limits comparability across studies. Sample sizes were typically small, and follow-up outcomes were rarely reported, restricting the interpretation of the long-term clinical efficacy of these devices. Conclusions Knee exoskeletons demonstrate promising short-term functional and biomechanical benefits and are increasingly feasible within structured rehabilitation settings. While current evidence is limited by small sample sizes and brief follow-up periods, emerging data suggest potential for enhancing task-specific gait training. Larger, well-designed trials are needed to determine sustained clinical impact and integration into routine rehabilitation practice.

The shoulder girdle is one of the most complex components of the upper limb, and when coupled with the arm, modeling and prediction become even more challenging. To address this, we propose a modular motion-planning framework that explicitly decouples the shoulder girdle from the arm. This separation simplifies modeling, enhances interpretability, and improves adaptability for rehabilitation scenarios. By independently targeting scapular dynamics, the framework enables better generalization of motion prediction across subjects. For human-motion modeling, a Transformer-based deep learning architecture is employed to capture nonlinear dependencies between joint angles and scapular motion. The model accepts joint-specific features as input and predicts shoulder girdle configurations, which are then integrated with arm trajectories to reconstruct complete upper limb motion. In the exoskeleton-mapping stage, a machine learning framework translates predicted human motion into the configuration space of a 6- Degree of freedom (DOF) rehabilitation exoskeleton. This ensures that generated trajectories are physically realizable and clinically suitable. By isolating human-motion prediction from robot mapping, the framework remains modular, scalable, and resilient to subject-specific variability, making it ideal for personalized rehabilitation. The methodology was evaluated by comparing Transformer-based predictions with both experimental data, reinforcement learning models and long short-term memory across multiple rehabilitation tasks. Quantitative analyses included statistical measures (F1-Score, ANOVA, T-test) and kinematic error metrics (RMSE, DTW). Results demonstrated that the Transformer model achieved higher accuracy and better temporal alignment with experimental trajectories. Combining shoulder-arm separation with Transformer-based learning provides an effective and clinically relevant solution for generating human-like motion in upper limb rehabilitation.

In rehabilitation robotics, control of energy flow is not merely a stability requirement but a therapeutic tool that shapes the quality and safety of human-robot interaction (HRI). The precise modulation of potential and kinetic energy within the coupled human-robot system governs how assistance is provided, how disturbances are rejected, and how patient effort is encouraged. Energy shaping approaches enable the controller to sculpt an artificial energy landscape anchored at a prescribed reference posture, so that restorative torques emerge naturally from the gradient of the shaped potential and disturbance-rich interactions are regulated within bounded, passive operating limits. This study presents a novel deep learning-based energy shaping framework for torque control in a three-degree-of-freedom (DOF) ankle rehabilitation robot. The proposed method is rooted in port-Hamiltonian mechanics. It employs interconnection and damping assignment-passivity-based control (IDA-PBC) to shape the energy landscape of the system, promoting practical stability and safe patient interaction. To address the limitations of static or heuristic energy shaping, we introduce a physics-informed data-driven approach in which the potential energy function is dynamically constructed through an Attention-Augmented Fourier Neural Operator (AFNO). This architecture learns mappings from spatiotemporal sensor data, including joint kinematics and interaction torques, to optimal shaping parameters that define the control energy field. The control strategy was experimentally validated on an ankle rehabilitation robot with ten healthy subjects (eight male, two female, aged 25-43), performing controlled movements across dorsiflexion/plantarflexion, inversion/eversion, and abduction/adduction. Experimental data confirmed that the shaped potential energy fields successfully guided joint trajectories toward the prescribed reference posture under disturbance-rich interaction conditions, while maintaining passivity and minimizing unnecessary energy expenditure.

Age-related Motor Dysfunction will inevitably cause a reduction in physical movement capacity and life quality among the elderly, and raise the risk of falls. Lower robotic exoskeletons can offer assistance to the wearer, while most of them are rigid, bulky, and not suitable for daily use. Employing a pneumatic actuating mechanism, a novel soft lower limb robotic exoskeleton was developed in this paper. The pneumatic actuators were designed as airbags and assembled in a queue to form the main body of the soft exoskeleton. To correctly and timely control the exoskeleton to assist the user, a pair of plantar pressure sensors was embedded in the shoes to detect the moment of leg lifting and recognize the gait. The assistive and biomechanical performances of the exoskeleton were evaluated with healthy human subjects. Qualitative assessment of the developed exoskeleton was also conducted. Experimental results show that the developed soft exoskeleton can achieve an assistive efficiency of 5% similar to 10 % in different conditions, which is quite acceptable for assisting the daily activities of elderly people.

This study describes the development and implementation of a robotic perturbation-based training methodology for children with cerebral palsy (CP). A Robotic Perturbation Trainer (RPT) was engineered to deliver controlled multidirectional waist-pull perturbations during treadmill walking. The system integrates a cable-driven parallel manipulator with a treadmill and body-weight support to enable safe and repeatable application of external disturbances in the transverse plane. The methodology was implemented in two adolescents with CP (ages 16-17; spastic diplegic and spastic dyskinetic subtypes) using a 5-week protocol consisting of ten treadmill-based sessions. Training parameters, including perturbation magnitude (65-85 N) and treadmill speed (0.2-0.5 km/h), were progressively adjusted according to predefined ranges. A multimodal assessment framework was applied at baseline, mid-, and post-implementation stages. Functional mobility and balance were evaluated using standardized outcome measures (6-Minute Walk Test, 10-Meter Walk Test, Timed Up and Go Test, Berg Balance Scale, and Gross Motor Function Measure), and surface electromyography was used to quantify activation of the gluteus maximus and medius muscles. The proposed methodology: Delivers controlled transverse-plane perturbations during treadmill walking using a cable-driven robotic system Applies a structured, progressive perturbation training protocol under body-weight support conditions Integrates clinical assessment tools with EMG-based muscle activity analysis for multimodal evaluation.

The risk of falling represents a significant barrier preventing many older adults from engaging in mass sports and physical activity. Objective assessment of the fall risk using wearable technologies constitutes an essential support in safe sport participation through early detection of gait instability. This study proposes a hybrid Temporal Convolutional Network (TCN)-based framework for gait-based fall risk identification based on lower back-mounted inertial measurement unit (IMU) sensor data acquired during one-minute laboratory tests and three-day free-living recording. The proposed methodology integrates data-driven temporal modeling of raw inertial signals with clinically interpretable handcrafted gait features. Temporal dependencies are modeled using dilated convolutions, while subject-level predictions are obtained through statistical aggregation of window-level representations. The framework is evaluated using subject-wise cross-validation and demonstrates consistent discrimination between fallers and non-fallers.The proposed methodology:● Processes waveform-level gait dynamics and captures detailed gait dynamics from raw accelerometer signals from short accelerometer recordings using the TCN.● Aggregates window-level embeddings at the subject level using statistical descriptors.● Incorporates clinically interpretable gait features that describe spatiotemporal characteristics, inter-axis coordination, and time- and frequency-domain properties of walking.

Despite advances in robot-assisted rehabilitation, existing control strategies often lack real-time personalization to account for subject-specific physiological capabilities, limiting the effectiveness of assist-as-needed interventions. Moreover, there is a further need to account for the patient's physiological functional capacity (PFC) in the design of such assistive technology. Here, an attempt has been made to create an Assist-as-Needed (AAN) Controller that works in collaboration with a Smart Avatar. The Smart Avatar mimics the patient's capabilities by learning in real-time and informing the controller. To efficiently predict the subject's level of involvement, a Reinforcement Learning (RL) based Inverse Dynamics model has been designed. Additionally, the controller for the Avatar has been appended with an Energy Map for modifying the reference trajectories and making the system energy efficient. The human torque estimated by the Smart Avatar assists the Assist-as-Needed (AAN) controller in providing the optimum robot torque to guide the subject's wrist along the modified trajectories. The developed algorithm was validated on five healthy participants. The system achieved trajectory tracking errors in the range of 0.01-0.04 rad across wrist motions. Subject- and axis-dependent differences in interaction torque were examined using descriptive statistics and torque profiles, supporting the feasibility of adaptive assistance under the tested conditions.

This research work explores the conceptualization and evaluation of a hybrid microgrid that taps into the potential of solar photovoltaic systems, wind energy conversion systems, and battery energy storage systems. In practice, nonlinear loads pose power quality issues that significantly impact the performance of hybrid microgrids. To tackle these challenges, an adaptive multi-generalized integrator (MGI) filter is proposed. This filter extracts the fundamental signal from the distorted utility grid voltages. It features a pre-filter with a DC-off set rejection loop, effectively minimizing the DC-offsets from the distorted grid voltages. This facilitates the distribution of active power generated by the hybrid microgrid's sources while simultaneously addressing various power quality problems. In addition, a dynamic power management control approach enhances grid resilience by operating the hybrid microgrid in grid-connected and islanding modes. It provides uninterrupted and highquality power supply to consumers during grid outages. The performance of this filter is comprehensively assessed through numerical simulations in MATLAB (R)/Simulink (R) environment. A real-time laboratory prototype is developed with WAVECT (R) WUC300 FPGA controller, demonstrating the practical application of the proposed filter. It ensures the minimum DC-offsets of 0.02V, fast convergence, and minimum oscillations. The grid synchronization and power quality index of the grid currents simultaneously achieve a THD of 2.82 %, complying with the IEEE-1547 and IEEE-519 standards. Importantly, the proposed adaptive-MGI filter outperforms the conventional SOGI and LMF filters in terms of nonlinear load tracking capabilities, with a response speed of less than 200 mu s, thereby highlighting its practical relevance and importance.

To efficiently control the motion of the upper limb rehabilitation robot, it is necessary that correct joint torques are provided. The analytical methods to compute the torque from the inverse dynamics are complex and intricate. For this purpose, a data driven approach based on deep learning model is presented in this study. The algorithm learns the position, velocities, and acceleration of the joints for a given trajectory and predicts the torque required. The results show the efficacy of the proposed algorithm in predicting the joint torques for the given dynamic parameters. The results obtained from this study can be further used in control of the upper limb rehabilitation robot.

Individuals with cerebral palsy (CP) experience significant impairments in lower limb mobility, which severely limit their daily activities and overall quality of life. Robotic exoskeletons have emerged as a cutting-edge solution to assist in the rehabilitation of individuals with CP by improving their motor functions. This systematic review, conducted following PRISMA guidelines, critically evaluates lower limb robotic exoskeletons specifically designed for individuals with CP, focusing on their design, rehabilitation interfaces, and clinical effectiveness. The review includes research papers published between 2010 and 2024, analyzing 30 lower limb exoskeletons reported in 57 papers. We analyze each exoskeleton, focusing on its technological features, user experience, and clinical outcomes. Notably, we identify a trend in which researchers are increasingly adapting exoskeleton functions to the specific needs of individual users, facilitating personalized rehabilitation approaches. Additionally, we highlight critical gaps in current research, such as the lack of sufficient long-term evaluations and studies assessing sustained therapeutic impacts. While ease of use remains crucial for these devices, there is a pressing need for user-friendly designs that promote prolonged engagement and adherence to therapy. This comprehensive review of existing gait rehabilitation exoskeleton technologies aimed to inform future design and application, ultimately contributing to the development of devices that better address the needs of individuals with CP and enhance their motor functions and quality of life.

The integration of renewable energy sources and distributed energy resources (DERs) has driven the evolution of modernized nested microgrids, enhancing resilience and flexibility in power distribution systems. Grid-following (GFL) and grid-forming (GFM) inverters are central to these systems, with GFL units emulating current sources challenged by uncertain grid impedance, and GFM units emulating voltage sources required to adapt to dynamic load variations. Mode transitions introduce instability through multi-loop control interactions. This work presents a comprehensive dynamical stability analysis of GFL and GFM inverters in nested microgrids, supported by advanced control strategies addressing dynamic response limitations, sensor dependencies, filter fluctuations, and controller complexities. An eigenvalue-based framework identifies dominant oscillatory modes, while online adaptation mitigates disturbances to preserve closed-loop performance. Time-evolution modeling of observables enables enhanced real-time monitoring. A blockchain-enabled decentralized framework ensures secure, transparent, and automated stability actions. Hardware-in-the-loop (HIL) experiments on a modified IEEE 123-node test feeder demonstrate a total harmonic distortion (THD) of 1.75% under weak-grid conditions compared with 2.73%, 4.76%, 8.40%, and 2.2% for other approaches and 0.3% under grid-impedance variation and <0.3% under nonlinear loading. The proposed controller achieves 0.06% tracking error dynamics and 0.02% steady-state error, outperforming classical methods (0.32–0.87% and 0.17–0.38%, respectively), with a computational time of 29 ms. The blockchain layer, implemented on the Polygon network, achieved a measured throughput of 1,572 transactions/s, an average block time of 2.3 s, and transaction fees below $0.01 USD, enabling rapid, economical, and scalable peer-to-peer stability service execution.

Robotic exoskeletons are being increasingly used in clinics for the treatment of medicable disabilities. These exoskeletons, which closely couple with patients’ limbs, need to move in harmony with the endoskeleton motions. To achieve coordination, exoskeletons should be transparent; in other words, they should not interfere with natural human motion or their underlying coordination strategies. Transparency can be achieved through a bio-inspired exoskeleton design and also by implementing appropriate force control methods to maneuver exoskeleton motions. A new hybrid active-passive Gait Exoskeleton-Assisted Rehabilitation (GEAR) robot is presented here for the rehabilitation of lower limb disabilities. The GEAR robot is designed to enhance transparency incorporating a flexible hip joint and a biomimetic knee joint. The proposed GEAR robot also integrates a Remote Centered Motion (RCM) based passive mechanism to support torso and pelvic motions in two planes and features actuated exoskeleton legs in the sagittal plane for treadmill-assisted walking. The exoskeleton legs are actuated at their hip and knee joints using backdrivable actuators. To provide a natural walking experience, the hip joints of the exoskeleton legs offer two passive degrees of freedom in the frontal and transverse planes in addition to the actuated sagittal plane motion. The biomimetic design of the exoskeleton knee joint ensures alignment with the human anatomical knee joint by closely tracking the latter’s instantaneous center of rotation (ICR). To evaluate GEAR robot’s transparency, a comparative study was conducted, involving three healthy subjects. The participants walked freely on a treadmill and then with the GEAR robot operated first in a completely backdrivable (i.e., passive) mode and subsequently in an active mode. The sEMG data collected during these experiments were analyzed to assess robot’s transparency.

Anterior cruciate ligament (ACL) injuries are among the most common and clinically significant knee disorders, and accurate detection from MRI remains essential for timely intervention. Recent advances in deep learning have shown promising results in analyzing MRI scans for ACL diagnosis. At the same time, self-supervised learning (SSL) has emerged as a powerful strategy to learn robust feature representations from unlabeled data. In this work, we evaluate the use of the Bootstrap Your Own Latent (BYOL) method for pretraining a ResNet-18 encoder, which is subsequently employed for ACL tear detection. Specifically, the encoder is first pretrained on unlabeled MRI scans to generate feature embeddings. These embeddings are then transferred to a downstream classifier to assess their effectiveness in improving classification accuracy. • Leveraging self-supervised learning to extract informative features from unlabeled knee MRI data using the BYOL framework. • Employing a pretrained ResNet-18 encoder to enhance feature representation for anterior cruciate ligament tear detection.

Recent advances in robotics and artificial intelligence have highlighted the potential for the integration of computational intelligence in enhancing the functionality and adaptability of robotic systems, particularly in rehabilitation. Designing robotic exoskeletons for the lower limb rehabilitation of post-stroke patients requires frequent adjustments to accommodate individual differences in leg anatomy. This complex engineering challenge necessitates a deep understanding of human physiology, robotics, and optimization to develop adaptive robotic systems and also to swiftly quantify the required adjustments and implement them for each patient. The conventional approaches, which mostly rely on heuristics and manual tuning, often struggle to achieve optimal results. This paper presents a novel method that integrates a genetic algorithm with a deep learning approach to generate a gait trajectory of the ankle joint from a six-bar linkage mechanism of fixed dimensions. Later, using the same approach, the inverse kinematics solution for this mechanism is also devised whereby, the set of the link dimensions of the six-bar linkage mechanism is obtained for the given gait trajectory of an individual to achieve customization. We simulated the kinematic behavior of the six-bar linkage mechanism within defined mechanical constraints and utilized the generated data for training a feedforward neural network and long short-term memory models. The proposed model, when trained, can produce accurate lengths for the desired gait trajectories in the sagittal plane and vice versa, which further validates our proposed approach for inverse kinematics solution. Moreover, to evaluate the efficiency of deep learning models, we have conducted an extensive error-based, comparative, and sensitivity analysis using different performance indices. The results highlight the potential of the proposed deep-learning-driven approach in the design analysis of gait rehabilitation robots.

In rehabilitation robotics, optimizing energy consumption and high interaction forces is essential to prevent unnecessary muscle fatigue and excessive joint loading as they often cause an inefficient trajectory planning and disrupt natural movement patterns. Stroke patients frequently exhibit asymmetrical muscle activation and impaired neuromuscular coordination, making it necessary to design a system that adapts to their specific motor limitations with energy-efficient and excessive torque control. This study presents a reinforcement learning-based trajectory optimization framework for a 3-DOF ankle rehabilitation robot, integrating musculoskeletal modeling, transactive energy and real-time physiological feedback to generate adaptive rehabilitation trajectories. The methodology utilizes electromyography (EMG) signals from key ankle muscles and joint reaction forces to refine movement patterns to ensure biomechanical efficiency. The methodology is validated using data from ten stroke patients, demonstrating its potential to enhance rehabilitation effectiveness by promoting more natural, efficient, and physiologically accurate movement trajectories.

Over the years, there has been a surge in the usage of robotic exoskeletons in clinics to treat patients with neurological impairments and other brain acquired injuries. For a comfortable and safe human-robot interaction, a harmony is required between the motion of exoskeleton and naturalistic movements of the user. To support natural coordination, exoskeletons must be transparent, ensuring natural human movements to proceed unimpeded. A biomimetic exoskeleton design, coupled with effective force control strategies, can help achieve transparency in motion. An upper limb rehabilitation robot is presented in this study to aid in the rehabilitation of upper limb disabilities. The proposed upper limb rehabilitation robot (ULRR) is designed to improve the transparency, incorporating shoulder, elbow, and wrist joints. The proposed ULRR provides six actuated degrees of freedom (DOF) out of which three are associated with the shoulder joint, one with the elbow joint, and two DOF for the wrist joint. The telescopic features at the upper and lower arm of the ULRR allow a close alignment between the elbow joint of the robot and the user. To evaluate the transparency of the ULRR, two comparative studies were conducted with ten healthy subjects. Firstly, the tests were conducted for the motions of the shoulder joint only in all three DOF individually. Later, a trajectory from the active daily life (ADL) was given as a reference and the ULRR was operated in patient-in-charge and robot-in-charge control modes. The force data collected during these experiments were analyzed to assess the transparency of the ULRR for human-robot interaction.

Integrating solar photovoltaic (PV), wind, and battery storage (BS) systems into the grid introduces significant power quality (PQ) challenges. In particular, the intermittent nature of solar PV and wind energy systems (WES), combined with nonlinear loads, can lead to grid instability. As a result, maintaining a reliable and high-quality power supply to consumers becomes a substantial challenge. This article presents two novel solutions to address these challenges: the anti-windup mixed-order generalized integrator (AWMOGI) and the adaptive delay operation period filter (ADOPF). The AWMOGI effectively extracts fundamental components from distorted utility grid voltages, minimizing DC offset and reducing steady-state oscillations. Meanwhile, the ADOPF swiftly captures positive sequence voltage components by introducing a two-sampling-interval delay in the d-q coordinate system. The hierarchical-loop controller parameters are optimized using a customized Harris Hawks Optimization (HHO) algorithm, which enhances error reduction and facilitates rapid establishment of reference values for DC-link voltage loop control. An adaptive dynamic power management scheme also enables the microgrid (MG) to manage surplus power from MG sources, reducing stress on the BS and ensuring DC-link voltage stability. The experimental results demonstrate significant improvements in handling utility grid abnormalities, dynamic loads, and intermittent microgrid source operation. The experimental results show a zero DC-off set and steady-state errors, a fast computational time of 35e-5s, and the THD of grid currents of 3.16%, which complies with the IEEE 519 standards.

The application of robotic devices in rehabilitation is proliferating. Such devices' mechanism design, actuation, and control strategy are essential for effective and successful rehabilitation treatment. This paper investigates the effectiveness of a self-aligning mechanism for a multi-DOFs (Degrees of Freedom) rehabilitation robot. The actuation is provided by lightweight albeit powerful Pneumatic Muscle Actuators (PMA). Although the mechanism design and the actuation system provide a safe, secure, and efficient platform for rehabilitation, they increase the complexity of the system modeling and, subsequently, the control system's design. Furthermore, the mechanism has three active and five passive DOFs, which further increase the intricacies of system identification. Hence, this paper presents an autodidactic approach to identify the system dynamics using the Koopman operator. The learned operator is then integrated with the Nonlinear Model Predictive Controller (NMPC) to guide the robot along the predefined path while adapting to the nonlinear dynamics of the physical human-robot interaction. Finally, the rehabilitation robot and the control scheme were experimentally validated with healthy human subjects. The results demonstrate that the NMPC controller could successfully manipulate the gait rehabilitation robot with the subject to achieve the desired orientation during the entire gait cycle.

This paper presents an adaptive Assist-as-Needed (AAN) control framework for a shoulder rehabilitation robot enhanced by a Virtual Biomechanical Shoulder Robot Model (VBSRM) and an online stiffness adaptation module. The proposed system adapts support levels dynamically based on user interaction and motor effort, ensuring both safety and active participation during rehabilitation training. The VBSRM is first calibrated to each user's anthropometric dimensions and used to estimate joint torques during movement. Assistance coefficient $\beta $ (t), is then dynamically updated based on the user's measured interaction force, increasing support when effort is low and reducing it when active participation improves. The AAN controller features a gradient-based online stiffness learning mechanism, enabling the system to dynamically adjust joint stiffness during movement based on tracking errors and interaction forces. This dual adaptation enhances rehabilitation training individualization. Experiments with ten healthy subjects performing an Activity of Daily Living (ADL) task task demonstrated improved torque profiles, effective stiffness modulation, and reduced trajectory tracking errors in active and passive modes. The proposed control scheme demonstrates high potential for patient-specific, responsive rehabilitation training in neurorehabilitation and post-stroke therapy.