With the recent progress in underwater robotics, autonomous underwater vehicles (AUVs) are increasing their mission capacity and range of applications operating in proximity to underwater infrastructure and closer to the wave zone for data collection and environmental monitoring. Future applications will necessitate more complex articulated systems that will be tasked to operate with increased safety, efficiency, and capacity. In this article, we present a novel motion planning framework for enabling, for the first time, safe autonomous navigation and active perception in high-dimensional articulated AUVs, employed with multiple arbitrary-configured sensors, in complex dynamic environments that are prone to uncertainties, disturbances, and currents. Inspired by previous work, we expand on a locally optimal and empirically safe pipeline designed for holonomic AUVs and enable it to handle articulated AUVs with potentially non-trivial dynamics that are unknown to the planner. We test the proposed methodologies with challenging simulated cases using an underwater snake robot (USR) with realistic dynamics and a generic underwater vehicle-manipulator system (UVMS) comprising of a holonomic base employed with a manipulator. Our results showcase the effectiveness of the proposed technique in producing efficient and safe real-time solutions, while discussing current limitations, computational bottlenecks, and potential solutions.
Physical human–robot interaction involves a tightly coupled cognitive interaction with the user. The emotional state of the user is therefore a key factor influencing engagement, motivation, and long-term acceptance. This review explores recent developments in emotion-aware technologies, particularly facial expression recognition and multimodal affect detection, and their potential to make assistive robots more adaptive and human-centred. By analysing current methodologies, emerging trends, and ethical considerations, we show how the integration of emotional intelligence can transform robotic assistance, shifting from reactive mechanisms to genuinely empathetic, socio-psychological, and context-aware assistive partners.
IntroductionOpen-set recognition capability is becoming necessary for collaborative robots because actual shop-floor dynamics do not always correspond to the fixed action classes used during training. Misclassification of unknown human motions may lead to inappropriate robot responses, whereas reliable rejection enables safer supervisory control.MethodsAn open-set human action recognition framework is developed by combining a spatial Convolutional Neural Network encoder with a Temporal Convolutional Network-based temporal aggregation module and an unknown-action rejection gate based on open-set scoring. The framework is evaluated on the NTU RGB+D 120 dataset using cross-subject and cross-view protocols with held-out action classes.ResultsThe proposed framework achieves 93.20% and 91.00% top-1 accuracy under cross-subject and cross-view evaluation, respectively, with corresponding Macro-F1 scores of 88.40% and 86.20%. For open-set detection, it achieves an AUROC of 0.938 and an AUPR of 0.750, with a 5.00% false acceptance rate for unknown actions and a 91.82% true known-action acceptance rate. The system operates at 15.0 ms per window and 66 fps on an NVIDIA Jetson Orin NX 16 GB, with a model size of 42 MB.DiscussionThe results demonstrate that calibrated unknown-action rejection can improve safety-aware human action recognition while retaining practical real-time performance for collaborative robot workspaces.
Robot-assisted spinal surgery has evolved from bone-mounted trajectory guidance to integrated navigation and, in selected research systems, constrained task execution. This evolution has been accompanied by rapid commercial and academic activity, but regulatory status, technical capability, and clinical effectiveness are often conflated. Such conflation can overstate both the maturity of emerging platforms and the certainty of their patient-level benefits. This structured narrative review was developed from a reproducible PubMed search, targeted citation chasing, Crossref metadata verification, and model-specific regulatory checks. Evidence was organized by source class and by outcome domain rather than by manufacturer chronology alone. Established platforms were separated from early clinical and preclinical systems, and regulatory clearance or registration was treated as distinct from evidence of comparative effectiveness. The most consistent evidence concerns pedicle-screw placement accuracy. Comparative studies and evidence syntheses generally favor robot-assisted placement over freehand or fluoroscopy-guided techniques, although the magnitude and consistency of benefit vary by platform, comparator, grading system, and surgeon experience. Radiation exposure may be reduced in some workflows, but acquisition protocols and reporting units are heterogeneous. Evidence for operative time, complications, revision surgery, patient-reported outcomes, and cost-effectiveness remains inconsistent or sparse. The principal contribution of this review is an evidence-maturity framework that links each technical capability to the strongest available clinical evidence and explicitly separates premarket regulatory decisions from claims of clinical superiority. Robot-assisted systems should currently be understood as surgeon-controlled enabling technologies. Future progress will depend less on adding degrees of freedom than on demonstrating reproducible patient benefit, safe task autonomy, interoperable data pipelines, and value across health-care settings.
