Frontiers in Robotics and AI最新文献

Background: Brain-Machine Interfaces (BMIs) increasingly mediate human interaction with assistive systems, yet remain sensitive to internal cognitive divergence. Subtle shifts in user intention-due to fatigue, overload, or schema conflict-may affect system reliability. While decoding accuracy has improved, most systems still lack mechanisms to communicate internal uncertainty or reasoning dynamics in real time.

Objective: We present NECAP-Interaction, a neuro-symbolic architecture that explores the potential of symbolic feedback to support real-time human-AI alignment. The framework aims to improve neuroergonomic transparency by integrating symbolic trace generation into the BMI control pipeline.

Methods: All evaluations were conducted using high-fidelity synthetic agents across three simulation tasks (motor control, visual attention, cognitive inhibition). NECAP-Interaction generates symbolic descriptors of epistemic shifts, supporting co-adaptive human-system communication. We report trace clarity, response latency, and symbolic coverage using structured replay analysis and interpretability metrics.

Results: NECAP-Interaction anticipated behavioral divergence up to 2.3 ± 0.4 s before error onset and maintained over 90% symbolic trace interpretability across uncertainty tiers. In simulated overlays, symbolic feedback improved user comprehension of system states and reduced latency to trust collapse compared to baseline architectures (CNN, RNN).

Conclusion: Cognitive interpretability is not merely a technical concern-it is a design priority. By embedding symbolic introspection into BMI workflows, NECAP-Interaction supports user transparency and co-regulated interaction in cognitively demanding contexts. These findings contribute to the development of human-centered neurotechnologies where explainability is experienced in real time.

Objective: Registering a preoperative 3D model of an organ with its actual anatomy viewed from an intraoperative video is a fundamental challenge in computer-assisted surgery, especially for surgical augmented reality. To address this, we present a benchmark of state-of-the-art deep learning point-cloud registration methods, offering a transparent evaluation of their generalizability to surgical scenarios and establishing a robust guideline for developing advanced non-rigid algorithms.

Methods: We systematically evaluate traditional and deep learning GMM-based, correspondence-based, correspondence-free, matching-based, and liver-specific point cloud registration approaches on two surgical datasets: a deformed IRCAD liver set and DePoll dataset. We also propose our complete-to-partial point cloud registration framework that leverages keypoint extraction, overlap estimation, and a Transformer-based architecture, culminating in competitive registration results.

Results: Experimental evaluations on deformed IRCAD tests reveal that most deep learning methods achieve good registration performances with TRE<10 mm, MAE(R) < 4 and MAE(t)<5 mm. On DePoll, however, performance drops dramatically due to the large deformations.

Conclusion: In conclusion, deep-learning rigid registration methods remain reliable under small deformations and varying partiality but lose accuracy when faced with severe non-rigid changes. To overcome this, future work should focus on building non-rigid registration architectures that preserve the strengths of self-, cross-attention and overlap modules while enhancing correspondence estimation to handle large deformations in laparoscopic surgery.

Recent advances in artificial intelligence (AI) have attracted significant attention due to AI's ability to solve complex problems and the rapid development of learning algorithms and computational power. Among the many AI techniques, transformers stand out for their flexible architectures and high computational capacity. Unlike traditional neural networks, transformers use mechanisms such as self-attention with positional encoding, which enable them to effectively capture long-range dependencies in sequential and spatial data. This paper presents a comparison of various deep Q-learning algorithms and proposes two original techniques that use self-attention into deep Q-learning. The first technique is structured self-attention with deep Q-learning, and the second uses multi-head attention with deep Q-learning. These methods are compared with different types of deep Q-learning and other temporal techniques in uncertain tasks, such as throwing objects to unknown targets. The performance of these algorithms is evaluated in a simplified environment, where the task involves throwing a ball using a robotic arm manipulator. This setup provides a controlled scenario to analyze the algorithms' efficiency and effectiveness in solving dynamic control problems. Additional constraints are introduced to evaluate performance under more complex conditions, such as a joint lock or the presence of obstacles like a wall near the robot or the target. The output of the algorithm includes the correct joint configurations and trajectories for throwing to unknown target positions. The use of multi-head attention has enhanced the robot's ability to prioritize and interact with critical environmental features. The paper also includes a comparison of temporal difference algorithms to address constraints on the robot's joints. These algorithms are capable of finding solutions within the limitations of existing hardware, enabling robots to interact intelligently and autonomously with their environment.

This paper discusses the functional advantages of the Selection-Broadcast Cycle structure proposed by Global Workspace Theory (GWT), inspired by human consciousness, particularly focusing on its applicability to artificial intelligence and robotics in dynamic, real-time scenarios. While previous studies often examined the Selection and Broadcast processes independently, this research emphasizes their combined cyclic structure and the resulting benefits for real-time cognitive systems. Specifically, the paper identifies three primary benefits: Dynamic Thinking Adaptation, Experience-Based Adaptation, and Immediate Real-Time Adaptation. This work highlights GWT's potential as a cognitive architecture suitable for sophisticated decision-making and adaptive performance in unsupervised, dynamic environments. It suggests new directions for the development and implementation of robust, general-purpose AI and robotics systems capable of managing complex, real-world tasks.

In recent years, several studies that analyze and interpret fish behavioral patterns in aquaculture settings have been published. Understanding how the fish react and respond to various scenarios and treatments can help provide insight and knowledge on how to achieve sustainable and efficient aquaculture production. Many of these research efforts have been conducted in land based tanks as this allows for closer and more continuous monitoring of the fish than what is possible at commercial facilities, essentially improving data quality and hence the possible insights to gain from these. However, most experimental tanks are closed-loop environments that are not particularly similar to commercial production units, as a consequence the results obtained in these systems are not directly transferable to industrial setups. Moreover, tank monitoring in such trials is often done using a single or a limited selection of different observation modes, which may not be sufficient to capture the full dynamics of fish responses. The present study seeks to address these challenges by developing the Cyber-Enhanced tank environment for aquaculture research. This concept features a tank environment setup to simulate the prevailing conditions in aquaculture units, mimicking natural light conditions, hiding sensors and other systems to reduce impacts on the fish and potential collisions, and using a tank color known to stimulate positive welfare in farmed fish. The tank was equipped with a novel sensor suite for high-fidelity detection and monitoring of fish behaviors based on a combination of an event camera, a scanning imaging sonar and conventional cameras. This innovative concept represents a step towards conducting experimental setups that are both more realistic in that conditions resemble those in commercial facilities and that uses a multi-modal sensor approach to capture details in fish responses in behaviors. The setup will be used as a basis for future fish responses experiments monitoring experiments in intensive aquaculture tanks.

Introduction: Socially assistive robots (SARs) have shown promise in pediatric healthcare by helping children manage the stress and anxiety associated with medical procedures. However, limited research exists on the specific robot behaviors that are most effective in reducing negative emotions in children during stressful interventions. This study aimed to compare the effectiveness of two emotional support strategies provided by a SAR during a vaccination event: internal emotion regulation through a guided breathing exercise and external emotion regulation via motivational speech and physical comfort (hugging). Additionally, we compared the effects of active and passive participation in the two SAR interventions.

Methods: A field study was conducted during annual group vaccination days, involving 225 children aged 8-12 years. Emotional and behavioral outcomes, including anxiety, fear, trust, and willingness to engage with the robot, were measured using self-report questionnaires.

Results: Results indicated that while girls reported higher levels of fear and anxiety than boys, active participation in the SAR intervention led to greater reductions in fear and anxiety, particularly among girls. Additionally, active hugging enhanced both engagement and trust, with girls showing a stronger response to such a physical comfort intervention.

Discussion: These findings indicate that, within the constraints of this study, SAR interventions were associated with reduced negative emotions in children during vaccinations, with active participation and physical comfort being particularly impactful for emotional support. This study offers valuable insights into optimizing SAR interventions in pediatric healthcare.

Embodied intelligent systems build upon the foundations of behavioral robotics and classical cognitive architectures. They integrate multimodal perception, world modeling, and adaptive control to support closed-loop interaction in dynamic and uncertain environments. Recent breakthroughs in Multimodal Large Models (MLMs) and World Models (WMs) are profoundly transforming this field, providing the tools to achieve its long-envisioned capabilities of semantic understanding and robust generalization. Targeting the central challenge of how modern MLMs and WMs jointly advance embodied intelligence, this review provides a comprehensive overview across key dimensions, including multimodal perception, cross-modal alignment, adaptive decision-making, and Sim-to-Real transfer. Furthermore, we systematize these components into a three-stage theoretical framework termed "Dynamic Perception-Task Adaptation (DP-TA)". This framework integrates multimodal perception modeling, causally driven world state prediction, and semantically guided strategy optimization, establishing a comprehensive "perception-modeling-decision" loop. To support this, we introduce a "Feature-Conditioned Modal Alignment (F-CMA)" mechanism to enhance cross-modal fusion under task constraints.

In macro world, the robots are programmed machines, engineered to perform repetitive and often specialized tasks. Scaling down the size of a robot by a billionth of a meter gives a nano robot. The primary driving force behind the advent of micro and nano robots have always been the domain of medical technology and these robots are popularly known as nano bio-robots. The current review shines a light on these bio-robots in all their facets, encompassing both top-down and bottom-up fabrication approaches to their associated challenges followed by ethical approvals. The study describes in detail the synthesis techniques of nano bio-robot along with required actuation mechanism of the bio-robots. Further, in this paper, how a nano biorobotic drug-delivery system (NDDS) can deliver the drugs in a controlled way to the targeted site of the host in contrast to conventional drug administration is discussed. The paper also reviews and summarizes the administration pathways of these bio-robots in the human body and their efficacy in reducing various disorders. Overall, it can be said that the integration of nano robots with bio-concept presents distinct advantages and possesses significant promises for many applications.

As part of the robotics technologies required for In-situ resource utilization (ISRU), the development of cargo rovers for transporting resources is needed. However, these cargo rovers have unique technical challenges that differ from conventional exploration rovers, including the need to traverse rough terrains with their varying mass due to transporting payloads. Moreover, research addressing these challenges has been limited, and the relevant technologies have not been fully established. To address these challenges, this paper proposes a parametric model for estimating wheel slippage. The model is formulated as a function of four input parameters: slope angle, rover heading angle, payload mass, and wheel angular velocity, and is applicable to resource-transporting rovers with varying mass. Additionally, the use of a parametric model reduces computational load, which offers advantages for onboard implementation. The proposed estimation model was quantitatively evaluated by comparing datasets obtained from multi-body dynamics analysis. This paper also introduces a new traversability assessment model which incorporates the proposed slip estimation model. We demonstrated the proposed model by integrating it into a sampling based motion planning. The simulation result of the motion planning show that the planner with our model can generate safer motions and enables the rover to reach the target regardless of the cargo payload.