Tendon-driven continuum manipulators can perform tasks in confined environments due to their flexibility and curvilinearity, especially in minimally invasive surgeries. However, the friction along tendons and tendon slack present challenges to their motion control. This work proposes a trajectory tracking controller based on adaptive fuzzy sliding mode control (AFSMC) for the tendon-driven continuum manipulators. It consists of a sliding mode control (SMC) law with two groups of adaptive fuzzy subcontrollers. The first one is utilized to estimate and compensate for friction forces along tendons. The second one adapts the switching terms of SMC to alleviate the chattering phenomenon and enhance control robustness. To prevent tendon slack, an antagonistic strategy along with the AFSMC controller is adopted to allocate driving forces. Simulation and experiment studies have been conducted to investigate the efficacy of the proposed controller. In free space experiments, the AFSMC controller generates an average root-mean-square error (RMSE) of 0.42% compared with 0.90% of the SMC controller. In the case of a 50 g load, the proposed controller reduces the average RMSE to 1.47% compared with 4.29% of the SMC controller. These experimental results demonstrate that the proposed AFSMC controller has high control accuracy, robustness, and reduced chattering.

Simultaneous Localization and Mapping (SLAM) is widely used to solve the localization problem of unmanned devices such as robots. However, in degraded environments, the accuracy of SLAM is greatly reduced due to the lack of constrained features. In this article, we propose a deep learning-based adaptive compensation strategy for sensors. First, we create a dataset dedicated to training a degradation detection model, which contains coordinate data of particle swarms with different distributional features, and endow the model with degradation detection capability through supervised learning. Second, we design a lightweight network model with short computation time and good accuracy for real-time degradation detection tasks. Finally, an adaptive compensation strategy for sensors based on the degree of degradation is designed, where the SLAM is able to assign different weights to the sensor information according to the degree of degradation given by the model, to adjust the contribution of different sensors in the pose optimization process. We demonstrate through simulation experiments and real experiments that the robustness of the improved SLAM in degraded environments is significantly enhanced, and the accuracy of localization and mapping are improved.

Hip joint moments during walking are the key foundation for hip exoskeleton assistance control. Most recent studies have shown estimating hip joint moments instantaneously offers a lot of advantages compared to generating assistive torque profiles based on gait estimation, such as simple sensor requirements and adaptability to variable walking speeds. However, existing joint moment estimation methods still suffer from a lack of personalization, leading to estimation accuracy degradation for new users. To address the challenges, this paper proposes a hip joint moment estimation method based on generalized moment features (GMF). A GMF generator is constructed to learn GMF of the joint moment which is invariant to individual variations while remaining decodable into joint moments through a dedicated decoder. Utilizing this well-featured representation, a GRU-based neural network is used to predict GMF with joint kinematics data, which can easily be acquired by hip exoskeleton encoders. The proposed estimation method achieves a root mean square error of 0.1180 ± 0.0021 Nm/kg under 28 walking speed conditions on a treadmill dataset, improved by 6.5% compared to the model without body parameter fusion, and by 8.3% for the conventional fusion model with body parameter. Furthermore, the proposed method was employed on a hip exoskeleton with only encoder sensors and achieved an average 20.5% metabolic reduction () for users compared to assist-off condition in level-ground walking.

While recent advancements in hybrid propulsion systems for bionic robotic fish—combining biomimetic mechanisms with classical vector thrusters—demonstrate enhanced locomotion capabilities and application potential, challenges remain in modeling the coupled dynamics of heterogeneous propulsion mechanisms. This paper presents a hybrid-drive robotic fish architecture that synergistically integrates pectoral-fin-mounted propellers with a caudal-fin-based propulsion system. A three-dimensional dynamical model is developed to characterize the coupled interactions between the dual propulsion modes, incorporating a hydrodynamic computation framework that accounts for propeller wake effects on caudal fin performance. Systematic experimental validation confirms the model’s fidelity through quantitative analysis of swimming performance metrics, including cruising speed, turning radius, and trajectory tracking. The results show that the proposed hybrid propulsion strategy can effectively improve the swimming performance of the robotic fish, and the model can effectively predict the motions such as speed, turning diameter, and trajectory of the robotic fish, which provides a new idea for the development of bionic robotic fish.

In recent years, the number of stroke patients worldwide has been steadily increasing, with approximately 70% of survivors experiencing upper limb dysfunction, particularly severe impairment of fine motor skills in the hand. This limitation significantly reduces patients’ ability to perform daily activities and increases the burden on both families and society. Existing hand rehabilitation exoskeletons suffer from issues such as complex structures, high production and usage costs, and limited application scenarios. This paper presents a flexible and portable hand rehabilitation robotic device based on the anatomical structure and movement characteristics of the human hand. First, a flexible exoskeleton glove based on underactuation is designed to accommodate various finger sizes. The portable device allows for rehabilitation in both hospital and home environments. Second, Adams simulation is used to verify the structural feasibility of the designed exoskeleton. Finally, device testing is performed on subjects to assess the assistive performance and motor dexterity of the hand exoskeleton using joint angle similarity tests, object grasping experiments, and force distribution tests. The experimental results show that the hand exoskeleton prototype can assist finger joints in achieving significant flexion and extension movements. Moreover, by adjusting the driving forces at each joint, it can stabilize the grasping of objects with different sizes, providing a high level of motion assistance in daily object grasping and finger joint movements. This study offers a practical and feasible technological path to reduce disability rates and improve the quality of life for patients with hand dysfunction following a stroke.

This study investigates the generalisation and explainability challenges of Robotic Foundation Models (RFMs) in industrial applications, using Octo as a representative case study. Motivated by the scarcity of domain-specific data and the need for safe evaluation environments, we adopt a simulation-first approach: instead of transitioning from simulation to real-world scenarios, we aim to adapt real-world-trained RFMs to synthetic, simulated environments — a critical step towards their safe and effective industrial deployment. While Octo promises zero-shot generalisation, our experiments reveal significant performance degradation when applied in simulation, despite minimal task and observation domain shifts. To explain this behaviour, we introduce a modified Grad-CAM technique that enables insight into Octo’s internal reasoning and focus areas. Our results highlight key limitations in Octo’s visual generalisation and language grounding capabilities under distribution shifts. We further identify architectural and benchmarking challenges across the broader RFM landscape. Based on our findings, we propose concrete guidelines for future RFM development, with an emphasis on explainability, modularity, and robust benchmarking — critical enablers for applying RFMs in safety-critical and data-scarce industrial environments.

Among the various soft actuators explored for robotic applications, the pneumatic muscle actuators (PMAs) stand out because of many advantages, such as compliant structures, high power-to-weight/volume ratios, and lightweight materials. Despite these advantages, their inherent nonlinearities and time-varying dynamics pose significant challenges for tracking control. To tackle this challenge, we present a robust control method that is structurally simple and computationally inexpensive. Such a method is comprised of an error transformation scheme, which is deeply explored to withstand model uncertainties to accomplish the output tracking with assigned accuracy, and a tuning function for relaxing requirements on the initial conditions. Experimental results of the PMA are presented to validate the concepts.

Currently, multi-UAV collision detection and avoidance is facing many challenges, such as navigating in cluttered environments with dynamic obstacles while equipped with low-cost perception devices having a limited field of view (FOV). To this end, we propose a communication-aided collision detection and avoidance method based on curriculum reinforcement learning (CRL). This method integrates perception and communication data to improve environmental understanding, allowing UAVs to handle potential collisions that may go unnoticed. Furthermore, given the challenges in policy learning caused by the substantial differences in scale between perception and communication data, we employ a two-stage training approach, which performs training with the network expanded from part to whole. In the first stage, we train a partial policy network in an obstacle-free environment for inter-UAV collision avoidance. In the second stage, the full network is trained in a complex environment with obstacles, enabling both inter-UAV collision avoidance and obstacle avoidance. Experiments with PX4 software-in-the-loop (SITL) simulations and real flights demonstrate that our method outperforms state-of-the-art baselines in terms of reliability of collision avoidance, including the DRL-based method and NH-ORCA (Non-Holonomic Optimal Reciprocal Collision Avoidance). Besides, the proposed method achieves zero-shot transfer from simulation to real-world environments that were never experienced during training.

Recent advancements in reinforcement learning (RL) and computational resources have demonstrated the efficacy of data-driven methodologies for robotic locomotion control and physical design optimization, providing a scalable alternative to traditional human-crafted design paradigms. However, existing co-design approaches face a critical challenge: the computational intractability of exploring high-dimensional design spaces, exacerbated by the resource-intensive nature of policy training and candidate design evaluations. To address this limitation, we propose an efficient co-adaptation framework for humanoid robot kinematics optimization. Building on a bi-level optimization architecture that jointly optimizes mechanical designs and control policies, our method achieves computational efficiency through two synergistic strategies: (1) a universal policy generalizable across design variations, and (2) a surrogate-assisted fitness evaluation mechanism. We implement the method with humanoid robot Kuafu, and by experimental results we demonstrate the proposed method effectively reduces the cost and the optimized design can achieve near-optimal performance.

Open tumor resection is one of the most commonly used treatments for malignant liver tumors. The ability to accurately locate the liver tumor during the operation is the key to the success of the operation. Intraoperative liver tumor localization remains challenging due to tissue deformation and intraoperative imaging limitations. This paper proposes a dual-constraint framework that synergistically integrates liver surface deformation and vascular biomechanical modeling to resolve this problem. Liver surface registration captures global deformation using a fast finite-element model (18 s), while vascular topology matching refines internal tumor displacement by enforcing correspondence between preoperative and intraoperative vessel trees. This synergistic strategy leverages both external and internal anatomical cues to achieve robust localization. Evaluated on 13 clinical cases, our method achieved sub-millimeter tumor localization accuracy (1.68±0.22 mm). Compared to single-constraint methods (LTLS: 2.04±0.26 mm; LTBV: 2.23±0.31 mm), our approach reduced error by 24%–37% without increasing runtime. This clinically efficient method shows promise for improving intraoperative guidance during liver tumor ablation.

As human–robot interaction (HRI) technology advances, dexterous robotic hands are playing a dual role—serving both as tools for manipulation and as channels for non-verbal communication. While much of the existing research emphasizes improving grasping and structural dexterity, the semantic dimension of gestures and its impact on user experience has been relatively overlooked. Studies from HRI and cognitive psychology consistently show that the naturalness and cognitive empathy of gestures significantly influence user trust, satisfaction, and engagement. This shift reflects a broader transition from mechanically driven designs toward cognitively empathic interactions — robots’​ ability to infer human affect, intent, and social context to generate appropriate nonverbal responses. In this paper, we argue that large language models (LLMs) enable a paradigm shift in gesture control — from rule-based execution to semantic-driven, context-aware generation. By leveraging LLMs and visual-language models, robots can interpret environmental and social cues, dynamically map emotions, and generate gestures aligned with human communication norms. We conducted a comprehensive review of research in dexterous hand mechanics, gesture semantics, and user experience evaluation, integrating insights from linguistics and cognitive science. Furthermore, we propose a closed-loop framework — “perception–cognition–generation–assessment” — to guide gesture design through iterative, multimodal feedback. This framework lays the conceptual foundation for building universal, adaptive, and emotionally intelligent gesture systems in future human–robot interaction.

Bio-syncretic robots represent a novel class of robots that integrate biological and artificial materials. These robots combine the high energy efficiency and environmental adaptability of biological tissues with the precise control and programmability of traditional robots, making them a focal point in the field of robotics. This paper reviews the latest research progress in bio-syncretic robots. Initially, we classify and introduce bio-syncretic robots from the perspective of structural design, which incorporates both biological and artificial materials. Subsequently, we provide a detailed discussion of their fabrication techniques and control methodologies. Finally, to facilitate broader applications of bio-syncretic robots, this paper explores their potential applications and future development prospects.

Micro Air Vehicle (MAV) swarms are often constrained by limited onboard processing capabilities and payload capacity, restricting the use of sophisticated localization systems. Lightweight ultra-wideband (UWB) ranging techniques are commonly used to estimate inter-vehicle distances, but they do not provide local bearing information—essential for precise relative positioning. Inspired by bat echolocation in low-visibility environments, we propose an acoustic-enhanced method for local bearing estimation designed for low-cost MAVs. Our approach leverages ambient acoustic signals naturally emitted by a target MAV in flight, combined with UWB distance measurements. The acoustic data is processed using the Frequency-Sliding Generalized Cross-Correlation (FS-GCC) method, enhanced with our analytical formulation that compensates for inter-channel switching delays in asynchronous, high-frequency sampling. This enables accurate Time Difference of Arrival (TDOA) estimation, even with compact microphone arrays. These TDOA values, along with known microphone geometry and UWB data, are integrated into our geometric model to estimate the bearing of the target MAV. We validate our approach in a controlled indoor hall across two experimental scenarios: static-bearing estimation, where the target MAV hovers at predefined angular positions (0°, ±30°, ±45°, ±60°), and dynamic-bearing estimation, where it flies across angles at varying velocities. The results show that our method yields reliable TDOA measurements compared to classical and machine learning baselines, and produces accurate bearing estimates in both static and dynamic settings. This demonstrates the feasibility of our low-cost acoustic-enhanced solution for local bearing estimation in MAV swarms, supporting improved relative navigation and decentralized perception in GPS-denied or visually degraded environments.