Miniaturization has emerged as a major technological trajectory in robotic surgery, encompassing single-port systems, flexible endoscopic platforms, capsule robotics, and microrobots designed to reduce surgical trauma and improve procedural precision. Despite rapid growth in this area, no previous bibliometric study has comprehensively mapped miniaturization as an integrated technological and clinical evolution across robotic surgery. Therefore, this study aimed to systematically analyze the global research landscape, collaboration patterns, thematic evolution, and emerging technological trends in miniaturized robotic surgery using advanced bibliometric and science-mapping approaches. A bibliometric analysis of 1,775 original articles indexed in the Elsevier Scopus database between 1996 and 2026 was conducted using Bibliometrix, VOSviewer, and CiteSpace to evaluate publication trends, collaboration networks, thematic evolution, citation bursts, and emerging technological trajectories. The literature demonstrated rapid expansion, with an annual growth rate of 17.1%, involving 6,293 authors across 492 sources and an international collaboration rate of 21.93%. The United States dominated scientific productivity and citation impact, followed by China, South Korea, Italy, and Germany. Thematic and cluster analyses identified robotic surgery, single-port systems, minimally invasive surgery, and medical robotics as the principal research domains. Citation-burst and trending-topic analyses revealed a temporal transition from capsule endoscopy, image-guided systems, and flexible robotics toward clinically deployable single-port robotic surgery, da Vinci SP platforms, partial nephrectomy, and postoperative pain. Highly cited studies emphasized continuum robotics, magnetic actuation, microrobotics, and biohybrid robotic systems, highlighting increasing convergence between robotic surgery, biomedical engineering, artificial intelligence, and nanotechnology. Miniaturization has evolved into a mature and rapidly expanding research domain characterized by strong technological convergence and increasing clinical translation. Current trends indicate a shift from feasibility-driven innovation toward intelligent, precision-oriented, and patient-centered robotic interventions.

The intelligent soft robotic gripper integrated with tactile sensors significantly enhances the robot’s execution capabilities in complex tasks, resolving critical shortcomings of traditional mechanical grippers—namely, fragile item breakage from rigid impacts, irregular object slippage, and inefficiency due to recognition errors. While electrical sensors (e.g., piezoresistive, capacitive) struggle with structural complexity, signal crosstalk, and environmental interference, optical waveguide tactile sensing offers superior sensitivity, rapid dynamics, and electromagnetic immunity. However, existing waveguide tactile systems face two key limitations: millimeter-scale waveguides cause beam divergence, limiting deformation sensitivity and complicating heterogeneous integration. Additionally, critical gaps remain in adaptive grasping control and contextual object recognition during manipulation. Herein, we present a soft robotic gripper integrated with slender elastic optical waveguide sensors (EOWS) and equipped with a closed-loop feedback control module to achieve intelligent grasping and object attribute recognition. The hand comprises three flexible silicone fingers, each finger seamlessly integrates three EOWS for multi-modal tactile sensing. These sensors exhibit high sensitivity to bending angle (0.273%/°), contact force (0.843%/N), and pressure (1.064%/N). Furthermore, a PID adaptive grasping control strategy and a long short-term memory (LSTM) deep learning algorithm are introduced to dynamically adjust the grasping force and intelligently recognize object attributes such as shape, size, and hardness, with accuracies exceeding 97% for each attribute. Ultimately, experimental validation via a smart fruit-sorting system highlights the platform’s potential for precision agriculture, intelligent logistics, and medical robotics, demonstrating robust, adaptive manipulation in real-world applications.We present a soft robotic gripper seamlessly integrated with slender multi-modal elastic optical waveguide sensors (EOWS) and equipped with an adaptive control module to achieve intelligent grasping and object attribute recognition. Experimental validation via a smart fruit-sorting system highlights the platform’s potential for precision agriculture, intelligent logistics, and medical robotics, demonstrating robust, adaptive manipulation in real-world applications

Soft magnetic actuators have gained significant interest for applications in minimally invasive medical robots, artificial muscles, soft robotic manipulators, and wearable bioelectronic interfaces, yet their functionality remains fundamentally limited by current magnetization strategies. To this end, a novel in-process printing and magnetization strategy with spatial and dynamic control of an external magnetic field during printing is developed to fabricate magnetorheological elastomers with fully customizable three-dimensional (3D) magnetization profiles. This method allows localized magnetic domain alignment in arbitrarily programmed orientations within a solid, enabling anisotropic actuation at micron to millimeter scales. The proposed method is highly sensitive to curing kinetics, material viscosity, and magnet positioning, which are characterized theoretically, experimentally, and in simulation. Structures magnetized in this way offer robust strain-sensing, information-encoding, and bio-inspired heterogeneous actuation capabilities. Demonstrations highlight this versatility, including a dragonfly with oppositely magnetized wings for tunable resonant actuation, an octopus-inspired swimmer whose magnetized legs reproduce aquatic locomotion, and a serpentine catheter with high degrees of freedom across 6 magnetic nodes. Together, these advances establish a versatile platform for designing magnetically responsive systems that couple programmable anisotropic actuation with biological complexity.

Hemodialysis demand is rising as populations age and the chronic kidney disease burden increases, yet dialysis units face persistent workforce constraints and substantial increases in cognitive and physical workloads, making workload reduction an urgent priority. We propose a tripartite collaboration model in which medical staff, artificial intelligence agents, and robots redesign hemodialysis workflows at the task level. Artificial intelligence agents support non-physical work through three coordinated modules: "Eye" integrates and visualizes multimodal data from dialysis machines, electronic health records, laboratories, and home or wearable monitoring to highlight early signals of deterioration; "Brain" uses machine learning to predict complications such as intradialytic hypotension and to support optimization of dry-weight estimation, anemia and chronic kidney disease-mineral and bone disorder management, and prescription trade-offs through scenario simulation; and "Language", based on large language models, drafts structured session summaries and plain-language explanations anchored to verified data, with clinician review to mitigate hallucinations and omissions. Robots reduce physical workload through equipment preparation, transport, and environmental maintenance, and may extend to reproducible vascular access surveillance using robotic ultrasound and, in the longer term, assisted cannulation. Clinicians/medical staff remain accountable for goal setting, value-laden decisions, communication, and authorization of automated outputs and actions. We also summarize governance requirements-interoperability, human factors evaluation, privacy and cybersecurity, and staged deployment starting from low-risk, verifiable functions. By delegating routine cognitive and physical work while preserving human responsibility and relational care, the model may enable more proactive, patient-centered hemodialysis and support sustainable staffing and workload reduction.

In the critical stage of digital and intelligent transformation of stomatology, artificial intelligence (AI) technology has shown significant advantages in diagnosis, treatment planning and other fields. However, limited by the nature of disembodied intelligence, the clinical execution end still faces the "last centimeter" bottleneck such as weak environmental perception and lack of dynamic interaction. As a new AI paradigm emphasizing the closed loop of "perception-decision-action", embodied intelligence provides a new theoretical direction and technical path to break through the execution limitations of existing digital technologies. This paper sorts out the technical core of embodied intelligence and its adaptation logic to oral clinical scenarios. From the perspective of different evaluation dimensions, it compares the conventional medical robot autonomy classification with intelligence-oriented frameworks, and further introduces the embodied intelligence grading system in the context of oral healthcare, systematically describes the development direction of core technologies such as visuo-tactile fusion perception, oral-specific world model, and simulation-to-reality (Sim2Real) simulation training, prospects its application scenarios in subspecialties such as oral implantology, prosthodontics, endodontics, and orthodontics, and puts forward expert suggestions on core issues such as data islands, regulatory ethics, and technical boundaries faced by the current field, so as to provide a reference for the rational development and standardized application of oral embodied intelligence.

Abstract This paper presents an intelligent flexible surgical platform that synergistically integrates high fidelity optical sensing multimodal human robot interaction and artificial intelligence driven decision making aligning closely with the paradigms of intelligent manufacturing in medical robotics. Designed for minimally invasive interventions in confined anatomical spaces the system features a dual channel flexible endoscope capable of 1920 by 1080 native resolution imaging providing robust visual input for real time semantic understanding via the EndoARSS framework built upon DINOv2. The optical data stream supports multi task learning architectures such as SMA and TESLA modules enabling accurate surgical activity recognition and tissue segmentation under complex intraoperative conditions. Coupled with a variable stiffness continuum manipulator with diameter no greater than 12 millimeters and bending angle exceeding 120 degrees and low latency gesture based control under 80 milliseconds the platform realizes a closed loop perception decision action workflow analogous to smart manufacturing systems. Safety is ensured through force threshold monitoring with warning triggered below 3 newtons and automatic shutdown above 5 newtons and electromagnetic compatibility compliant hardware design. By embedding optical sensing at the core of its intelligent architecture this work demonstrates how advanced vision technologies and artificial intelligence can transform surgical robots into adaptive context aware agents offering a compelling case study of intelligent manufacturing principles applied to next generation medical devices.

Magnetic navigation of photoacoustic/ultrasound catheters via vision-ultrasound fusion servo control for embodied medical robots

BackgroundJapan has been promoting the use of medical and long-term care robots to reduce the workload of healthcare professionals. In home-visit nursing, where the number of older adults and patients with dementia is increasing, robots may provide benefits, such as supporting infection control and physically demanding care. However, implementing these technologies may also generate new ethical challenges.ObjectiveTo clarify home-visiting nurses' perceptions and the ethical issues regarding the introduction of medical and caregiving robots into home-visiting nursing field in Japan.MethodsA mixed-methods design was employed. A questionnaire survey was distributed to one nurse from each of 1012 home-visit nursing stations in Prefecture A. Descriptive statistics were used for quantitative analysis. To gain deeper insights, interviews were conducted with eight nurses who consented to participate.Ethical considerationsThis study was approved by the institutional review board (Approval No.: 2022N-026). Participation was voluntary, and informed consent was obtained. Data were anonymized to ensure confidentiality.Results and discussionThe questionnaire response rate was 17%. Simple tabulation showed that visiting nurses had positive expectations regarding the introduction of medical care robots but lacked confidence in their specific usefulness and safety. Only "disinfection of equipment" was identified as a task that could be fully shifted to robots. Twenty-six tasks, including "rehabilitation," "massage," "bathing and shower care," and "relaxation," were considered shareable between nurses and robots. Fifteen tasks-such as "suctioning," "enema and fecal disimpaction," "postmortem care," and "explanations and communication with family"-were regarded as nurse-exclusive. Interview findings suggest that medical and nursing care robots may help reduce workload and supplement care for individuals living alone or requiring intensive support. However, key challenges include device safety and reliability, user and family acceptance, privacy protection, and implementation costs. Ethical consideration and clinical effectiveness must be carefully evaluated after clarifying target users, purpose, and scope of intervention.

The incorporation of Artificial Intelligence (AI) into medical robotics has transformed contemporary healthcare by improving precision, efficiency, and personalization in clinical treatments. This paper offers a thorough examination of AI-driven medical robots, emphasizing their transformative capabilities in diagnosis, surgery, rehabilitation, and patient care. Artificial intelligence algorithms, especially those utilizing machine learning and deep learning frameworks, empower medical robots to execute intricate tasks, including image-guided surgery, self-navigating, making decisions in real time, and adaptive learning. Surgical robots integrated with AI enhance minimally invasive treatments by providing exceptional precision and minimizing patient trauma, whereas diagnostic robots facilitate early illness identification through pattern recognition in imaging and genetic data. Rehabilitation robots, equipped with AI, provide personalized therapy by continuously assessing and adjusting to the patient's advancement. Moreover, socially helpful robots employ natural language processing (NLP) and affective computing to aid elderly and disabled patients via interactive care. Notwithstanding these gains, problems endure, encompassing ethical dilemmas, data privacy issues, regulatory adherence, and the necessity for rigorous validation in clinical environments. The paper examines the present state of AI-driven medical robotic systems, assesses ongoing clinical trials, and considers future trajectories, highlighting the imperative for cross-disciplinary cooperation among engineers, data scientists, physicians, and legislators. This study objectively evaluates the features and limitations of AI-driven medical robots, highlighting their significance as essential instruments in advancing precision medicine and transforming the global healthcare landscape through intelligent automation and improved patient-centred solutions.

PurposePrevious studies have explored the impact of medical robots on patient care, but few studies have looked at the impact of information delivery by medical robots on patient information compliance.Patients and MethodsIn this study, 290 subjects were recruited on a professional data collection platform from November 21 to December 10, 2024 to conduct scenario experiments to explore the effect of medical robot information transmission on cancer patients’ information compliance and its mechanism.ResultsThe experimental results show that medical robot information transmission has a significant impact on patient information compliance. Among them, patients’ preference for emotional information transmission by medical robots was higher than that for professional information transmission. At the same time, we found that information processing fluency and perceived psychological stress have a chain mediating effect on medical robot and patient information compliance.ConclusionThis study further expands the extension of information processing theory, strengthens the role of emotional factors and cognitive factors in human-computer interaction, and provides a new program for effective clinical nursing.

Simulation-based training is a critical component of robotic surgical education. While virtual reality platforms are well established for basic skills acquisition, they lack the ability to replicate tissue handling required for procedural simulation. Synthetic models have emerged as a promising alternative; however, formal validation is required prior to their integration into training curricula. This study aimed to establish the face and content validity of the International Medical Robotics Academy (IMRA) Surgical model for robotic transabdominal preperitoneal (TAPP) inguinal hernia repair. A prospective content validity study was conducted using an international expert panel. Surgeons with experience in robotic TAPP repair and prior exposure to the IMRA model were recruited. A 48-item survey instrument was developed and organized across domains including anatomical fidelity, procedural relevance, haptic properties, and educational utility. Items were rated on a 4-point Likert scale. Item-level content validity indices (I-CVI) were calculated, with a threshold of ≥ 0.78 for retention. The scale-level content validity index (S-CVI/Ave) was computed, with ≥ 0.90 considered acceptable. Face validity was assessed using five global rating items (1-10 scale). Ten expert surgeons from three countries participated. The overall face validity score was 8.78/10, with all domains exceeding 8.0, including educational value (9.40) and procedural relevance (8.90). The S-CVI/Ave was 0.929. Of 48 items, 44 (91.7%) achieved I-CVI ≥ 0.78. Four items did not meet the threshold, primarily related to advanced dissection steps and model-specific anatomical features. 90% of experts endorsed the model without reservations. The IMRA synthetic simulation model for robotic TAPP inguinal hernia repair demonstrates strong face and content validity. These findings support its integration into structured robotic training programs and provide a foundation for future studies evaluating construct and criterion validity.

Advances in additive manufacturing for medical robotics: A review

High-precision control of integrated joint modules is critical for medical robots, which demand extreme accuracy and stability. However, robustness is often compromised by rapidly time-varying uncertainties, including nonlinear friction, unmodeled parameters, electronic noise, and external disturbances, which collectively induce complex nonlinear dynamical behaviors. To address this challenge, a constrained second-order dynamical system based on angular constraints is first constructed. The Udwadia-Kalaba theory is then introduced to enforce desired constraints in deterministic systems. Furthermore, a novel recursive adaptive robust control method is proposed, explicitly incorporating multi-type uncertainties and nonlinear friction dynamics with discontinuous characteristics. Key advantages include eliminating the need for precise uncertainty boundaries and guaranteeing asymptotic error convergence via a recursive adaptive design. This ensures strict enforcement of system constraints despite nonlinearities. Simulations and experiments on a physical joint module and a real-time simulation control system high-speed controller prototype demonstrate significant reductions in tracking error and enhanced reliability under diverse uncertainties. The proposed method is validated in a joint module of medical robots, providing a generalizable framework to tackle nonlinear control problems.

AI-integrated multifunctional phase change e-skin: synergizing thermal management with multimodal sensing

This survey examines human involvement in safety-critical reinforcement learning (RL), focusing on safety enforcement rather than learning efficiency. Using a PRISMA-based systematic review, 100 studies published between 2010 and 2025 were analyzed across domains including autonomous driving, medical robotics, and power systems. The survey identifies a key gap: existing RL safety approaches rely heavily on algorithmic guarantees, which often fail under uncertainty, rare events, and high-stakes ethical trade-offs. To address this, I introduce the Human Safety Constraint Framework (HSCF), which formalizes human roles as preventive, corrective, advisory, and normative constraints. Case studies illustrate how human oversight complements algorithmic safeguards, mitigating residual risks and highlighting practical limitations such as cognitive load, latency, and trust calibration. The survey concludes that integrating human judgment as an explicit safety component is essential for robust, certifiable RL systems. Recommendations include developing formal models of human constraints, event-driven intervention strategies, and scalable hybrid architectures for real-world deployment.

Soft robotics introduces a new paradigm in biomedical engineering by offering inherent biocompatibility and mechanical compliance beyond the capabilities of conventional rigid systems. However, realizing translational potential requires a shift from material‐centric development toward strategically engineered design solutions. This perspective discusses the strategy‐driven evolution of soft actuators into robotic platforms, with emphasis on how architectural optimization addresses specific clinical challenges. Recent advances in implantable, surgical, and wearable soft robotic systems are comprehensively reviewed based on fundamental actuation mechanisms. By illustrating how material composition and system integration are deliberately tailored to diverse human systems, this article highlights mechanical interfacing, adaptive functionality, and durable long‐term operation. Collectively, the reviewed studies demonstrate that strategic design enables robust performance in unstructured physiological environments. Finally, progressive future directions for soft robotics are outlined to support intimate and adaptive interaction with neural and biological systems. These insights articulate a clear roadmap for translating soft robotics research into advanced human‐centered medical technologies that directly impact patient care and quality of life.

Intelligent sensing and measurement technologies for medical robotics and health monitoring

Motion optimization is a critical challenge in intelligent robotic systems, directly influencing efficiency, accuracy, and energy consumption. Traditional control and trajectory planning methods often rely on predefined models and heuristics, limiting their adaptability in dynamic and uncertain environments. This paper presents a machine learning–based approach for motion optimization in intelligent robotic systems, enabling adaptive, data-driven decision making for improved performance. The proposed framework integrates supervised and reinforcement learning techniques to optimize robot trajectories, joint coordination, and actuator control in real time. Sensor data, including position, velocity, and force feedback, are utilized to continuously learn and refine motion strategies. Experimental evaluations conducted on simulated and physical robotic platforms demonstrate significant improvements in trajectory smoothness, task completion time, and energy efficiency compared to conventional control methods. The results highlight the potential of machine learning to enhance autonomy and robustness in intelligent robotic systems, making the approach suitable for applications such as medical robotics, industrial automation, and assistive technologies.

Vision-based robotic triaxial tactile sensing provides superior spatial resolution and rich multimodal data. However, employing rigid CMOS imagers suffers from limitations in mechanical flexibility and large-area scalability. Here we present a large-area ultraflexible photoelectrical impedance tomography (PIT)-based imager that achieves high-fidelity triaxial tactile sensing. The 5-μm-thick PIT imager incorporates a quantum dots/metal-oxide heterojunction layer with 16 peripheral electrodes, significantly reducing interconnects complexity (pixel-to-interconnect ratio >80). The device exhibits a photo-to-dark-current ratio exceeding 10⁴ under ultraviolet illumination, resolves spatiotemporal features as fine as 1.5 mm, and can simultaneously image up to five occluded regions. Byintegrating a thinlight-scattering porous rubber and flexibleLEDs, triaxial force decoding is achieved through Gaussian photocurrent analysis.The systemachievesover a dynamic range of 80 kPa with a normal force sensitivity of 0.04 kPa⁻¹, a shear displacement resolution of 0.17 μm kPa⁻¹, and a topological recognition accuracy of 96.5%. We anticipate that this technology will enable advanced applications in industrial and humanoid robotics, medical and rehabilitation robotics, and wearable health monitoring and human-machine interaction systems.

Type of the article: Research ArticleAbstractThe integration of AI-enabled medical robots into the medical field has increased the potential risks to which patients may be exposed. To protect patients’ rights, this study aims to explore and analyze the need for mandatory insurance against civil liability of medical robots operating with AI technologies in the United Arab Emirates. Such insurance is intended to ensure adequate compensation, reinforce legal protection, and uphold confidence in medical practice, while also contributing to societal stability and supporting the growth of the insurance sector. The study employed a combination of descriptive and analytical methods. It concludes that smart medical robots are neither inanimate objects nor irrational beings. It recommends legislative regulations granting them digital legal personality under specific controls, recognizing their independent financial status, and enabling them to bear civil liability for actions causing harm. The study showed an upward trend in the number of insurance companies providing liability coverage for damages caused by AI-operated medical robots, increasing from two in 2020 to ten in 2025, and expected to rise further if full legal personality is granted. The research findings suggest amending the UAE Civil Transactions Code and the Medical Liability Law to codify civil liability provisions for autonomous smart medical robots and to mandate liability insurance. Furthermore, as insurers’ obligations depend on establishing the insured’s liability, UAE law should grant the injured party a direct right of action against the insurer.