Accurate 3D deformation control of deformable soft tissues is of paramount importance in robotic-assisted surgeries. Selecting optimal grasping points is a fundamental challenge, as the deformation behavior is highly dependent on the applied forces and their locations. This paper presents an efficient grasping point selection algorithm using optimization-based inverse finite element method for tissue manipulation tasks. We propose a method for the automatic identification of optimal grasping points that minimize feature or shape errors during deformation tasks. Specifically, we formulate the grasping task as a quadratic programming problem while considering the complex mechanical coupling within the tissue structure. Our method effectively accommodates both discrete key points and point clouds as input, and can simultaneously determine multiple optimal grasping points in one optimization process. We validate the proposed method in simulation on a tissue and liver model, demonstrating its feasibility and efficiency in various scenarios. Real-world experiments are conducted on a silicone liver phantom to further validate the effectiveness of our proposed method.

This research proposes an Internet of Things (IoT)-enabled adaptive robotic navigation framework tailored for smart campuses and urban mobility systems. It aims to overcome critical limitations in existing systems that rely on static data, lack real-time adaptability, and perform poorly in dynamic or adverse environments. The proposed system uniquely integrates heterogeneous real-time data sources including traffic, obstacle, and weather captured from IoT sensors into a unified decision-making architecture. It combines a graph neural network for dynamic environmental modeling, a convolutional neural network for obstacle mapping, and a multilayer perceptron for weather-aware path assessment. A proximal policy optimization-based reinforcement learning (RL) controller then computes continuous control actions. A novel multi-objective reward function adaptively adjusts priorities between travel time, energy efficiency, collision risk, and terrain stability based on the current IoT context, enabling fine-grained, scenario-aware optimization. The system is deployed on resource-constrained edge hardware (Jetson Nano), proving its feasibility for real-time embedded applications. Simulations across diverse scenarios including urban traffic congestion, dynamic obstacle handling, and adverse weather demonstrate 95% navigation accuracy, 98% obstacle detection precision, and near-optimal route selection. The framework sustains real-time operation with 10 Hz decision throughput and sub-300 ms latency, outperforming traditional static and rule-based systems while sustaining over 92% performance consistency under adverse weather. This work introduces a first-of-its-kind modular framework that fuses IoT sensory data, adaptive RL control, and edge deployment for robust, efficient navigation. It establishes a scalable baseline for real-world autonomous mobility in smart city ecosystems.

Effectively controlling systems governed by partial differential equations (PDEs) is crucial in several fields of applied sciences and engineering. These systems usually yield significant challenges to conventional control schemes due to their nonlinear dynamics, partial observability, high-dimensionality once discretized, distributed nature, and the requirement for low-latency feedback control. Reinforcement learning (RL), particularly deep RL (DRL), has recently emerged as a promising control paradigm for such systems, demonstrating exceptional capabilities in managing high-dimensional, nonlinear dynamics. However, DRL faces challenges, including sample inefficiency, robustness issues, and an overall lack of interpretability. To address these challenges, we propose a data-efficient, interpretable, and scalable Dyna-style model-based RL framework specifically tailored for PDE control. Our approach integrates Sparse Identification of Nonlinear Dynamics with Control within an Autoencoder-based dimensionality reduction scheme for PDE states and actions (AE+SINDy-C). This combination enables fast rollouts with significantly fewer environment interactions while providing an interpretable latent space representation of the PDE dynamics, facilitating insight into the control process. We validate our method on two PDE problems describing fluid flows—namely, the 1D Burgers equation and 2D Navier–Stokes equations—comparing it against a model-free baseline. Our extensive analysis highlights improved sample efficiency, stability, and interpretability in controlling complex PDE systems.

Most exoskeletons are designed with the shoulder joint’s instantaneous center of rotation (ICR) in mind as a fixed joint, often also known as the center of the shoulder joint. In fact, shoulder ICR changes during shoulder abduction–adduction and flexion–extension. Abduction–adduction causes the ICR to move in the frontal plane, which is caused by the joint movement of the shoulder joint, including depressed elevation and horizontal translation, while the flexion–extension movement of the sagittal plane produces the shoulder extension movement. If the change in shoulder ICoR movements is not compensated for in the exoskeleton design, they can create discomfort and pain for the robot’s wearer. Although conventional exoskeletons typically treat the shoulder joint as a three degree of freedom spherical joint, this study incorporates a more sophisticated understanding of shoulder kinematics. The developed scapulohumeral rhythm compensation mechanism successfully compensates for shoulder joint motion, with simulation results confirming kinematics that closely match ergonomic shoulder movement patterns. First, the complex kinematics of the shoulder joint are analyzed. To meet the demand for mismatch compensation, a shoulder exoskeleton based on a winding mechanism is designed. A mismatch compensation model is established, and theoretical analysis and simulation verify that the designed shoulder exoskeleton has a mismatch compensation function. While solving the mismatch problem, the human–machine coupling model is established through OpenSim software. The simulation results show that the designed exoskeleton has a good assisting effect from the perspective of muscle force generation and shoulder torque.

Grasp detection is a significant research direction in the field of robotics. Traditional analysis methods typically require prior knowledge of the object parameters, limiting grasp detection to structured environments and resulting in suboptimal performance. In recent years, the generative convolutional neural network (GCNN) has gained increasing attention, but they suffer from issues such as insufficient feature extraction capabilities and redundant noise. Therefore, we proposed an improved method for the GCNN, aimed at enabling fast and accurate grasp detection. First, a two-dimensional (2D) Gaussian kernel was introduced to re-encode grasp quality to address the issue of false positives in grasp rectangular metrics, emphasizing high-quality grasp poses near the central point. Additionally, to address the insufficient feature extraction capabilities of the shallow network, a receptive field module was added at the neck to enhance the network’s ability to extract distinctive features. Furthermore, the rich feature information in the decoding phase often contains redundant noise. To address this, we introduced a global-local feature fusion module to suppress noise and enhance features, enabling the model to focus more on target information. Finally, relevant evaluation experiments were conducted on public grasping datasets, including Cornell, Jacquard, and GraspNet-1 Billion, as well as in real-world robotic grasping scenarios. All results showed that the proposed method performs excellently in both prediction accuracy and inference speed and is practically feasible for robotic grasping.

Robotic manufacturing systems offer significant advantages, including increased flexibility and reduced costs. However, challenges in trajectory planning, error compensation, and system integration hinder their broader application in additive manufacturing. To address these issues, this paper proposes a generalized non-parametric trajectory planning method tailored for robotic additive manufacturing. The proposed trajectory planner incorporates chord error and speed continuity constraints and integrates the look-ahead planning with real-time interpolation in a parallel structure to ensure smooth transitions in the robot’s trajectory. Additionally, a real-time path tracking control method is introduced, combining RBF neural network-based dynamic feedforward control with visual servoing-based feedback control. This control strategy significantly improves the robot’s tracking accuracy, particularly for complex additive manufacturing paths that involve multiple short connected line segments and frequent speed variations. The effectiveness of the proposed methods is validated through experiments on a robotic additive manufacturing platform. The experimental results (line segment, circular arc segment, and continuous path tracking) show that the robot’s tracking error remains within $\pm$0.15 mm and $\pm 0.05^{\circ }$.

Collaborative design (co-design) is a team effort fostered through the creative involvement of all participants in co-creative collaboration (co-creation). This new approach to design as a creative social activity heightens the need to study the interpersonal aspects of creativity. Though co-creation has become widely used in recent years, few studies focus on its dynamics, which emerge from intense interactions created by the shared subjectivities of participants in an intersubjective environment. The management and enhancement of interpersonal factors can help create this shared environment by leading the process from personal to interpersonal creativity. Some of these interpersonal factors could be measured by observing the data of biosignals that are used as social cues, particularly if studied in comparison with the data of one of the partners of the social interaction, thanks to the synchrony rate between these datasets. This synchrony of biosignals related to shared behaviours can be associated with the interactive level dynamics that occur during co-creation in team of two (pairwork). This study presents the results of an experiment where biosignal synchrony results were compared to subjective feedback regarding the interactive level to understand the dynamics of the interaction. The results suggest the possibility of using the synchrony rate measured by the Damerau- Levenshtein distance (Ld) or dynamic time warping method (DTW) to approximate the dynamics of the interactive level in co-creative pairwork. This study will contribute to our understanding of the influence of the socio-cognitive process on interactions during co-creation to improve the co-creative design process.