The presence of abnormal muscle retraction and contraction during the rehabilitation training poses significant challenges for patients with lower limb dysfunction caused by stroke, as it hinders their ability to regain lower limb function. Therefore, it is vital to detect spasm force during the training to adjust the force generated by rehabilitation robots. This paper presents the design and development of a capacitive force/torque sensor for lower limb rehabilitation to detect muscle spasm. The double differential measurement method and a decoupling structure with built-in crossed roller bearing were adopted achieve high sensitivity and high accuracy. Calibration tests were conducted with a commercial torque and rotational speed sensor to verify the design and accuracy of capacitive force/torque sensor. Results show that the proposed sensor exhibits a linearity of 1.04%, a resolution of 0.05N·m and a maximum measurement error of approximately 2.07%. Lastly, experiments were conducted to assess spasm detection capabilities in a lower limb rehabilitation robot which has capacitive force/torque sensors installed on knee and hip joints.

Human gait exhibits stable contralateral coupling, making healthy-side motion a viable predictor for affected-limb kinematics. Leveraging this property, this study develops FusionTCN–Attention, a causality-preserving temporal model designed to forecast contralateral hip and knee trajectories from unilateral IMU measurements. The model integrates dilated temporal convolutions with a lightweight attention mechanism to enhance feature representation while maintaining strict real-time causality. Evaluated on twenty-one subjects, the method achieves hip and knee RMSEs of 5.71° and 7.43°, correlation coefficients over 0.9, and a deterministic phase lag of 14.56 ms, consistently outperforming conventional sequence models including Seq2Seq and causal Transformers. These results demonstrate that unilateral IMU sensing supports low-latency, stable prediction, thereby establishing a control-oriented methodological basis for unilateral prediction as a necessary engineering prerequisite for future hemiparetic exoskeleton applications.

Pneumatic separation can exhibit unstable performance when the feed composition fluctuates while operating parameters remain fixed. This work investigates a perception-informed airflow regulation approach, demonstrated on a representative fibrous–granular mixture case study. We propose LGDNet, a lightweight visual ratio estimation network (0.08 M parameters) built with Ghost-based operations and learned grouped channel convolution (LGCC), to estimate mixture composition from dense images. A dedicated 21-class dataset (0–100% in 5% increments) containing approximately 21,000 augmented images was constructed for training and evaluation. LGDNet achieves a Top-1 accuracy of 66.86%, an interval accuracy of 74.10% within a $\pm 5\%$ tolerance, and an MAE of 4.85, with an average inference latency of 28.25 ms per image under the unified benchmark settings. To assess the regulation mechanism, a coupled CFD–DEM simulation model of a zigzag air classifier was built and used to compare a regime-dependent airflow policy with a fixed-velocity baseline under representative prescribed inlet ratios. Under high impurity loading ($r=70\%$), the dynamic policy improves product purity by approximately 1.5 percentage points in simulation. Together, the real-image perception evaluation and the mechanism-level simulation study suggest the feasibility of using visual ratio estimation to inform airflow adjustment; broader generalization and further on-site validation on real equipment will be pursued in future work.

To address the challenge of autonomous process adaptation in non-standard components with continuously varying groove angles, this study proposes an intelligent decision-making framework based on Response Surface Methodology (RSM) for oscillation welding. Instead of solely identifying a single optimal parameter set, RSM is employed as a knowledge-modeling tool to reveal adaptive relationships between groove geometry and key welding parameters. A Central Composite Design (CCD) is utilized to establish predictive models for weld geometry under varying conditions: wire feed rate (8–12 m/min), travel speed (5–9 mm/s), travel angle (70–110°), oscillation amplitude (2–6 mm), dwell time (0.2–0.6 s), and groove angle (80–100°). The significance and adequacy of the models are validated through analysis of variance (ANOVA), demonstrating high predictive accuracy with all coefficients of determination (R²) exceeding 0.82. Furthermore, defect-aware physical constraints derived from the formation mechanism of bottom humping are incorporated into the optimization process, specifically restricting the travel angle to a push angle of 70–85° to ensure feasible and reliable decision outputs. Based on the established response surfaces, geometry-dependent parameter selection rules are derived to simultaneously optimize root penetration (target 8.5–10.5 mm) and sidewall fusion (>2.5 mm) for groove angles ranging from 80° to 100°. Experimental validation confirms that the proposed decision-making strategy achieves stable bead formation and defect-free fusion, demonstrating high quantitative reliability with root penetration prediction errors below 7% and bead width errors below 13%. This work bridges the gap between geometric perception and process control, providing a practical pathway toward intelligent and adaptive robotic welding of non-standard components.