Shigang Wang

Action quality assessment in vocational fitter training is constrained by limited expert-annotated data and insufficiently interpretable predictions. To achieve accurate scoring and rubric-aligned, auditable explanations with limited data, this study proposes RESC, a rubric-guided error–score consistency framework. RESC uses human pose as its primary visual evidence and converts an instructor rubric into error semantics and scoring constraints. It therefore predicts an overall score while producing rubric-aligned error diagnoses and model-derived itemized deduction contributions. We constructed Fitter-AQA, a dataset of 134 filing videos from 25 participants, and evaluated the framework using strict leave-one-subject-out cross-validation. RESC achieved a mean absolute error of 4.450, a Pearson linear correlation coefficient of 0.902, and a Spearman rank correlation coefficient of 0.852, corresponding to the lowest pooled scoring errors and highest pooled correlations among the evaluated methods. Error diagnosis reached a Macro-F1 of 0.770 and a macro-average balanced accuracy of 0.824. These results indicate stable cross-subject scoring under limited-data conditions and transform model predictions into an auditable score decomposition aligned with rubric semantics. RESC thus provides an accurate and interpretable approach to automated assessment in vocational skills training.

For humanoid robots to interact naturally with humans and seamlessly integrate into daily life, natural language serves as an essential communication medium. While recent advances in imitation learning have enabled robots to acquire complex motions through expert demonstration, traditional approaches often rely on rigid task specifications or single-modal inputs, limiting their ability to interpret high-level semantic instructions (e.g., natural language commands) or dynamically switch between actions. Directly translating natural language into executable control commands remains a significant challenge. To address this, we propose ToggleMimic, an end-to-end imitation learning framework that generates robotic motions from textual instructions, enabling language-driven multi-task control. In contrast to end-to-end methods that struggle with generalization or single-action models that lack flexibility, our ToggleMimic framework uniquely combines the following: (1) a two-stage policy distillation that efficiently bridges the sim-to-real gap, (2) a lightweight cross-attention mechanism for interpretable text-to-action mapping, and (3) a gating network that enhances robustness to linguistic variations. Extensive simulation and real-world experiments demonstrate the framework’s effectiveness, generalization capability, and robust text-guided control performance. This work establishes an efficient, interpretable, and scalable learning paradigm for cross-modal semantic-driven autonomous robot control.

Percutaneous puncture has become one of the most widely used minimally invasive techniques in clinical practice due to its advantages of low trauma, quick recovery and easy operation. However, incomplete needle tip movement, tissue barriers and complex distribution of sensitive organs make it difficult to balance puncture accuracy and safety. To this end, this paper proposes a new puncture path planning algorithm for flexible needles, which integrates gravitational guidance, bi-directional adaptive expansion, optimal node selection based on the A* algorithm, and path optimization strategies, with Bi-Rapid-Research Random Trees (Bi-RRTs) at its core, to significantly improve obstacle avoidance capability and computational efficiency. The simulation results of 2D and 3D complex scenes in MATLAB show that compared with the traditional RRT algorithm and Bi-RRT algorithm, the GBOPBi-RRT algorithm achieves significant advantages in terms of path length, computation time and node size. In particular, in the 3D environment, the GBOPBi-RRT algorithm shortens the planning path by 43.21% compared with RRT, 27.47% compared with RRT* and 30.33% compared with Bi-RRT, which provides a reliable solution for efficient planning of percutaneous puncture with flexible needles.

For the path planning of robots, a new robot path planning algorithm has been proposed. This algorithm is centered on the Bidirectional Rapidly-exploring Random Tree (Bi-RRT), integrating the gravitational guidance strategy and the optimal node selection strategy based on the A* algorithm. This significantly enhances the obstacle avoidance capability and computational efficiency. The simulation results in a twodimensional scenario using MATLAB show that, compared to the RRT, RRT*, and Bi-RRT, the GOBi-RRT achieves significant advantages in terms of path length, computational time, and node scale. It provides a reliable solution for the efficient planning of robot paths.

Legged robots are particularly well-suited for navigating terrains that are inaccessible to wheeled robots, which has led to a significant increase in their applications in recent years. However, many existing control approaches still struggle to achieve generalization in diverse and unstructured environments. In this work, we introduce the MSD training framework for general and agile locomotion of legged robots through a multi-skill distillation approach. First, expert policies are trained for different terrains. Then, using the DAgger algorithm, these expert skills are distilled into a single deployable policy. The proposed policy integrates depth images as exteroceptive observations, enabling agile navigation across diverse and unstructured terrains. Finally, deployment on the MagicDog-W robot validates the policy’s capability to achieve robust and flexible locomotion in challenging environments.

In the context of predicting pedestrian trajectories for indoor mobile robots, it is crucial to accurately measure the distance between indoor pedestrians and robots. This study aims to address this requirement by extracting pedestrians as regions of interest and mitigating issues related to inaccurate depth camera distance measurements and illumination conditions. To tackle these challenges, we focus on an improved version of the H-GrabCut image segmentation algorithm, which involves four steps for segmenting indoor pedestrians. Firstly, we leverage the YOLO-V5 object recognition algorithm to construct detection nodes. Next, we propose an enhanced BIL-MSRCR algorithm to enhance the edge details of pedestrians. Finally, we optimize the clustering features of the GrabCut algorithm by incorporating two-dimensional entropy, UV component distance, and LBP texture feature values. The experimental results demonstrate that our algorithm achieves a segmentation accuracy of 97.13% in both the INRIA dataset and real-world tests, outperforming alternative methods in terms of sensitivity, missegmentation rate, and intersection-over-union metrics. These experiments confirm the feasibility and practicality of our approach. The aforementioned findings will be utilized in the preliminary processing of indoor mobile robot pedestrian trajectory prediction and enable path planning based on the predicted results.

Lung cancer is the leading cause of cancer deaths worldwide. Although several lung cancer diagnostic methods are available for lung nodule biopsy, there are limitations in terms of accuracy, safety, and invasiveness. Transbronchial needle aspiration (TBNA) is a common method for diagnosing and treating lung cancer that involves a robot-assisted medical flexible needle moving along a curved three-dimensional trajectory, avoiding anatomical barriers to achieve clinically meaningful goals in humans. Inspired by the puncture angle between the needle tip and the vessel in venipuncture, we suggest that different orientations of the medical flexible needle puncture path affect the cost of the puncture trajectory and propose an effective puncture region based on the optimal puncture direction, which is a strategy based on imposing geometric constraints on the search space of the puncture direction, and based on this, we focused on the improved implementation of RCS*. Planning within the TBNA-based lung environment was performed using the rapidly exploring random tree (RRT), resolution-complete search (RCS), and RCS* (a resolution-optimal version of RCS) within an effective puncture region. The experimental results show that the optimal puncture direction corresponding to the lowest cost puncture trajectory is consistent among the three algorithms and RCS* is more efficient for planning. The experiments verified the feasibility and practicality of our proposed minimum puncture angle and puncture effective region and facilitated the study of the puncture direction of flexible needle puncture.