Frontiers in Robotics and AI最新文献

The bipedal wheel-legged robot combines the high energy efficiency of wheeled movement with the terrain adaptability of legged locomotion. However, achieving a smooth transition between these two heterogeneous motion modes within a unified control framework remains challenging. This study proposes a reinforcement learning control framework that integrates the Mixture of Experts (MoE) architecture. This approach employs a "divide and conquer" strategy by introducing a dynamic gating network and a Top-K sparse activation mechanism, which automatically allocates different motion modes to specific expert subnetworks, effectively decoupling conflicting gradients. Simulation results demonstrate that, compared to the single-network PPO method, the MoE-enhanced algorithm exhibits significant improvements in training stability and rewards. The learned policy successfully achieved smooth rolling on flat surfaces and transitioned to dynamic leg-lifting gaits when confronted with obstacles. In various test terrains, it showed a markedly higher success rate compared to the single-network PPO method.

This paper presents the design of a Brazilian robot, named Zequinha, for cultural and educational purposes in light of Human-Centered Artificial Intelligence challenges. Zequinha's development is a blend of art, robotics, and AI, evolving from MIDI-programmed animatronics to an autonomous entity integrating multiple local AIs. This shift to local processing inherently enhances privacy and governance, minimizing reliance on external APIs and enabling offline operability. The project's human-centered design approach is evident in its iterative methodology and its collaborative development with children. Zequinha promotes human wellbeing by enriching cultural mediation, engaging diverse audiences, and demonstrating potential in health and education. Moreover, the focus on local AI fosters responsible design and meaningful human-AI interaction, aiming to create a charismatic, safe, and useful robotic mediator.

This paper presents a conceptual framework for the design of personalized persuasive conversational agents to support positive behavior change. This paper leverages key theoretical models to understand the determinants of behavior change and explores how these models can inform the design of personalized conversational agents to enhance their effectiveness in healthcare interventions. The role of personalization in dialogue-based intervention is discussed, emphasizing the importance of adaptation to individual characteristics, preferences, and contexts. The potential of persuasive language generation is also examined, highlighting its ability to create more engaging and impactful behavior change strategies. Finally, the paper proposes a layered framework that explicitly links behavioral models, user personalization, and persuasive language generation, and discusses future research directions for integrating this framework in social robots' interventions for behavior change in healthcare.

Achieving robust, dexterous manipulation in unstructured environments remains a central challenge in robotics, particularly for continuous, contact-rich tasks like cleaning. While motion primitives can also be learned directly in full joint space, a compact, synergy-based representation provides a shared latent coordinate system that simplifies interpretation, modulation, and cross-task composition. We adopt a data-driven framework for representing and reproducing dexterous manipulation trajectories, using cleaning motions as a test bed. To model these movements, we combine Principal Component Analysis (PCA) with Probabilistic Movement Primitives (ProMPs), leveraging hand synergies. While the PCA and ProMP combination itself is established, our focus in this study, is on the cleaning use case and on the compositional generalization across tasks. PCA, applied in joint space, provides a compact, low-dimensional synergy space for coordinated finger movements, while the ProMPs encode the time-varying structure and variability of trajectories within this space. We first recorded a kinematic dataset of human cleaning motions with 20 degrees of freedom (DOF) haptic exoskeleton gloves across thirteen tasks and learn one ProMP per five selected training tasks in the PCA space. This dataset is then used as a basis to learn cleaning motions using the PCA + ProMPs. We demonstrate the ability of the learned primitives to reconstruct and reproduce kinematic patterns in simulation (Shadow Hand) and successfully deploy them on a physical robotic hand (Aeon Robotics). These results indicate that motion primitives, when grounded in synergy-informed coordinates, can generalize beyond grasping to encode and modulate contact-rich dexterous manipulation skills. Moreover, a library of the five task-specific ProMPs compositionally approximates trajectories from eight unseen cleaning tasks, with nearest-expert selection outperforming convex blends and Product-of-Experts combinations.

As artificial intelligence (AI) drives the development of next-generation robotic platforms and navigation systems that operate with increasing levels of autonomy in orthopedic and neurosurgical procedures, the methods by which human operators verify and validate these systems' operations become critically important. While significant effort has been spent on advancing technological capabilities and autonomy, comparatively little thought has been put into understanding how surgeons may effectively maintain oversight and assurance of these complex systems-despite retaining full legal and ethical responsibility for surgical outcomes. This mini-review synthesizes assurance mechanisms following the Sense-Think-Act framework: spatial intelligence (navigation and registration), cognitive assistance (AI-driven planning and adaptation), and physical operation (robot motion and force interaction). We highlight human-centered assurance as an opportunity to enable safe adoption of increasingly autonomous surgical systems. Finally, we outline essential research directions for developing assurance frameworks that scale with increasing autonomy while maintaining human responsibility and control in orthopedic and neurosurgical procedures.

ADMIRE (Analyzing Digitalized Human-Machine Interactions and Relationships) is a tool that was developed and tested as part of the Integrated Research Cluster. Its aim is to make explicit the implicit assumptions about humans and machines, as well as their potential and limitations. In this way, it provides a basis for structured, reflective research and development processes relating to human-machine interactions, as well as providing a starting point for ethical considerations in technology design. This article outlines the initial research and development approach and the insights gained from various research projects and application settings. We then trace this back to anthropology and the implicit images of humans and machines that determine the processes of research and development, and often prevent the implementation of 'technological solutions' to social problems. Here, we introduce the ADMIRE tool, along with its theoretical background and practical deployment. Finally, we reflect on the limitations of the tool itself and our experience to date.

Simulations of human Ia and Ib reflexes on a bio-inspired musculoskeletal robot driven by pneumatic artificial muscles (PAMs) offer a favorable option for counteracting disturbances in complex and dynamic work environments, providing a solution to the significant computational burdens that undermine its potential due to PAMs' inherent non-linearities. This research focuses on the simultaneous integration of human Ia and Ib reflexes (referred to as double-reflex) as countermeasures against physical disturbance in a musculoskeletal robot system driven by PAMs. The system's performance was examined, and implementation challenges were identified during experiments. Mechanisms were then applied to ensure the effective functioning of the integrated reflexes. Experimental results substantiated the effectiveness of the double-reflex system, highlighting its functionality within the robotic system. This investigation corroborates the viability of concurrently implementing Ia and Ib reflexes, providing a reference for the design of robotic reflex control systems. The study also offers some references based on the view of signal processing, regarding the possible functions of the human spinal cord that might be necessary to perform proper reflex actions.

This study develops a hierarchical learning and optimization framework that can learn and achieve well-coordinated multi-skill locomotion. The learned multi-skill policy can switch between skills automatically and naturally while tracking arbitrarily positioned goals and can recover from failures promptly. The proposed framework is composed of a deep reinforcement learning process and an optimization process. First, the contact pattern is incorporated into the reward terms to learn different types of gaits as separate policies without the need for any other references. Then, a higher-level policy is learned to generate weights for individual policies to compose multi-skill locomotion in a goal-tracking task setting. Skills are automatically and naturally switched according to the distance to the goal. The appropriate distances for skill switching are incorporated into the reward calculation for learning the high-level policy and are updated by an outer optimization loop as learning progresses. We first demonstrate successful multi-skill locomotion in comprehensive tasks on a simulated Unitree A1 quadruped robot. We also deploy the learned policy in the real world, showcasing trotting, bounding, galloping, and their natural transitions as the goal position changes. Moreover, the learned policy can react to unexpected failures at any time, perform prompt recovery, and successfully resume locomotion. Compared to baselines, our proposed approach achieves all the learned agile skills with improved learning performance, enabling smoother and more continuous skill transitions.