MagicGripper: A Multimodal Sensor-Integrated Gripper for Contact-Rich Robotic Manipulation

Contact-rich manipulation in unstructured environments demands precise, multimodal perception to enable robust and adaptive control. Vision-based tactile sensors (VBTSs) have emerged as an effective solution; however, conventional VBTSs often face challenges in achieving compact, multi-modal functionality due to hardware constraints and algorithmic complexity. In this work, we present MagicGripper, a multimodal sensor-integrated gripper designed for contact-rich robotic manipulation. Building on our prior design, MagicTac, we develop a compact variant, mini-MagicTac, which features a three-dimensional, multi-layered grid embedded in a soft elastomer. MagicGripper integrates mini-MagicTac, enabling high-resolution tactile feedback alongside proximity and visual sensing within a compact, gripper-compatible form factor. We conduct a thorough evaluation of mini-MagicTac's performance, demonstrating its capabilities in spatial resolution, contact localization, and force regression. We also assess its robustness across manufacturing variability, mechanical deformation, and sensing performance under real-world conditions. Furthermore, we validate the effectiveness of MagicGripper through three representative robotic tasks: a teleoperated assembly task, a contact-based alignment task, and an autonomous robotic grasping task. Across these experiments, MagicGripper exhibits reliable multimodal perception, accurate force estimation, and high adaptability to challenging manipulation scenarios. Our results highlight the potential of MagicGripper as a practical and versatile tool for embodied intelligence in complex, contact-rich environments.

Jigsaw-Puzzles: From Seeing to Understanding to Reasoning in Vision-Language Models

Spatial reasoning is a core component of human cognition, enabling individuals to perceive, comprehend, and interact with the physical world. It relies on a nuanced understanding of spatial structures and inter-object relationships, serving as the foundation for complex reasoning and decision-making. To investigate whether current vision-language models (VLMs) exhibit similar capability, we introduce Jigsaw-Puzzles, a novel benchmark consisting of 1,100 carefully curated real-world images with high spatial complexity. Based on this dataset, we design five tasks to rigorously evaluate VLMs' spatial perception, structural understanding, and reasoning capabilities, while deliberately minimizing reliance on domain-specific knowledge to better isolate and assess the general spatial reasoning capability. We conduct a comprehensive evaluation across 24 state-of-the-art VLMs. The results show that even the strongest model, Gemini-2.5-Pro, achieves only 77.14% overall accuracy and performs particularly poorly on the Order Generation task, with only 30.00% accuracy, far below the performance exceeding 90% achieved by human participants. This persistent gap underscores the need for continued progress, positioning Jigsaw-Puzzles as a challenging and diagnostic benchmark for advancing spatial reasoning research in VLMs.

A Dataset and Benchmarks for Deep Learning-Based Optical Microrobot Pose and Depth Perception

Optical microrobots, manipulated via optical tweezers (OT), have broad applications in biomedicine. However, reliable pose and depth perception remain fundamental challenges due to the transparent or low-contrast nature of the microrobots, as well as the noisy and dynamic conditions of the microscale environments in which they operate. An open dataset is crucial for enabling reproducible research, facilitating benchmarking, and accelerating the development of perception models tailored to microscale challenges. Standardised evaluation enables consistent comparison across algorithms, ensuring objective benchmarking and facilitating reproducible research. Here, we introduce the OpTical MicroRobot dataset (OTMR), the first publicly available dataset designed to support microrobot perception under the optical microscope. OTMR contains 232,881 images spanning 18 microrobot types and 176 distinct poses. We benchmarked the performance of eight deep learning models, including architectures derived via neural architecture search (NAS), on two key tasks: pose classification and depth regression. Results indicated that Vision Transformer (ViT) achieve the highest accuracy in pose classification, while depth regression benefits from deeper architectures. Additionally, increasing the size of the training dataset leads to substantial improvements across both tasks, highlighting OTMR's potential as a foundational resource for robust and generalisable microrobot perception in complex microscale environments.

Multi-Agent Systems for Robotic Autonomy with LLMs

Since the advent of Large Language Models (LLMs), various research based on such models have maintained significant academic attention and impact, especially in AI and robotics. In this paper, we propose a multi-agent framework with LLMs to construct an integrated system for robotic task analysis, mechanical design, and path generation. The framework includes three core agents: Task Analyst, Robot Designer, and Reinforcement Learning Designer. Outputs are formatted as multimodal results, such as code files or technical reports, for stronger understandability and usability. To evaluate generalizability comparatively, we conducted experiments with models from both GPT and DeepSeek. Results demonstrate that the proposed system can design feasible robots with control strategies when appropriate task inputs are provided, exhibiting substantial potential for enhancing the efficiency and accessibility of robotic system development in research and industrial applications.

Coarse-to-Fine Learning for Multi-Pipette Localisation in Robot-Assisted In Vivo Patch-Clamp

In vivo image-guided multi-pipette patch-clamp is essential for studying cellular interactions and network dynamics in neuroscience. However, current procedures mainly rely on manual expertise, which limits accessibility and scalability. Robotic automation presents a promising solution, but achieving precise real-time detection of multiple pipettes remains a challenge. Existing methods focus on ex vivo experiments or single pipette use, making them inadequate for in vivo multi-pipette scenarios. To address these challenges, we propose a heatmap-augmented coarse-to-fine learning technique to facilitate multi-pipette real-time localisation for robot-assisted in vivo patch-clamp. More specifically, we introduce a Generative Adversarial Network (GAN)-based module to remove background noise and enhance pipette visibility. We then introduce a two-stage Transformer model that starts with predicting the coarse heatmap of the pipette tips, followed by the fine-grained coordination regression module for precise tip localisation. To ensure robust training, we use the Hungarian algorithm for optimal matching between the predicted and actual locations of tips. Experimental results demonstrate that our method achieved > 98% accuracy within 10 {\mu}m, and > 89% accuracy within 5 {\mu}m for the localisation of multi-pipette tips. The average MSE is 2.52 {\mu}m.

Rethinking Bimanual Robotic Manipulation: Learning with Decoupled Interaction Framework

Bimanual robotic manipulation is an emerging and critical topic in the robotics community. Previous works primarily rely on integrated control models that take the perceptions and states of both arms as inputs to directly predict their actions. However, we think bimanual manipulation involves not only coordinated tasks but also various uncoordinated tasks that do not require explicit cooperation during execution, such as grasping objects with the closest hand, which integrated control frameworks ignore to consider due to their enforced cooperation in the early inputs. In this paper, we propose a novel decoupled interaction framework that considers the characteristics of different tasks in bimanual manipulation. The key insight of our framework is to assign an independent model to each arm to enhance the learning of uncoordinated tasks, while introducing a selective interaction module that adaptively learns weights from its own arm to improve the learning of coordinated tasks. Extensive experiments on seven tasks in the RoboTwin dataset demonstrate that: (1) Our framework achieves outstanding performance, with a 23.5% boost over the SOTA method. (2) Our framework is flexible and can be seamlessly integrated into existing methods. (3) Our framework can be effectively extended to multi-agent manipulation tasks, achieving a 28% boost over the integrated control SOTA. (4) The performance boost stems from the decoupled design itself, surpassing the SOTA by 16.5% in success rate with only 1/6 of the model size.

Design and Benchmarking of A Multi-Modality Sensor for Robotic Manipulation with GAN-Based Cross-Modality Interpretation

In this paper, we present the design and benchmark of an innovative sensor, ViTacTip, which fulfills the demand for advanced multi-modal sensing in a compact design. A notable feature of ViTacTip is its transparent skin, which incorporates a `see-through-skin' mechanism. This mechanism aims at capturing detailed object features upon contact, significantly improving both vision-based and proximity perception capabilities. In parallel, the biomimetic tips embedded in the sensor's skin are designed to amplify contact details, thus substantially augmenting tactile and derived force perception abilities. To demonstrate the multi-modal capabilities of ViTacTip, we developed a multi-task learning model that enables simultaneous recognition of hardness, material, and textures. To assess the functionality and validate the versatility of ViTacTip, we conducted extensive benchmarking experiments, including object recognition, contact point detection, pose regression, and grating identification. To facilitate seamless switching between various sensing modalities, we employed a Generative Adversarial Network (GAN)-based approach. This method enhances the applicability of the ViTacTip sensor across diverse environments by enabling cross-modality interpretation.

TCR-GPT: Integrating Autoregressive Model and Reinforcement Learning for T-Cell Receptor Repertoires Generation

T-cell receptors (TCRs) play a crucial role in the immune system by recognizing and binding to specific antigens presented by infected or cancerous cells. Understanding the sequence patterns of TCRs is essential for developing targeted immune therapies and designing effective vaccines. Language models, such as auto-regressive transformers, offer a powerful solution to this problem by learning the probability distributions of TCR repertoires, enabling the generation of new TCR sequences that inherit the underlying patterns of the repertoire. We introduce TCR-GPT, a probabilistic model built on a decoder-only transformer architecture, designed to uncover and replicate sequence patterns in TCR repertoires. TCR-GPT demonstrates an accuracy of 0.953 in inferring sequence probability distributions measured by Pearson correlation coefficient. Furthermore, by leveraging Reinforcement Learning(RL), we adapted the distribution of TCR sequences to generate TCRs capable of recognizing specific peptides, offering significant potential for advancing targeted immune therapies and vaccine development. With the efficacy of RL, fine-tuned pretrained TCR-GPT models demonstrated the ability to produce TCR repertoires likely to bind specific peptides, illustrating RL's efficiency in enhancing the model's adaptability to the probability distributions of biologically relevant TCR sequences.

CrystalTac: 3D-Printed Vision-Based Tactile Sensor Family through Rapid Monolithic Manufacturing Technique

Recently, vision-based tactile sensors (VBTSs) have gained popularity in robotics systems. The sensing mechanisms of most VBTSs can be categorised based on the type of tactile features they capture. Each category requires specific structural designs to convert physical contact into optical information. The complex architectures of VBTSs pose challenges for traditional manufacturing techniques in terms of design flexibility, cost-effectiveness, and quality stability. Previous research has shown that monolithic manufacturing using multi-material 3D printing technology can partially address these challenges. This study introduces the CrystalTac family, a series of VBTSs designed with a unique sensing mechanism and fabricated through rapid monolithic manufacturing. Case studies on CrystalTac-type sensors demonstrate their effective performance in tasks involving tactile perception, along with impressive cost-effectiveness and design flexibility. The CrystalTac family aims to highlight the potential of monolithic manufacturing in VBTS development and inspire further research in tactile sensing and manipulation.

Adaptive Extraction Network for Multivariate Long Sequence Time-Series Forecasting

Models employing CNN architecture have made significant progress in multivariate long sequence time-series forecasting (MLSTF), particularly in modeling local time series characteristics. However, during the MLSTF process, extracting the global time series patterns and understanding the correlations among different variables are highly significant. To address this challenge, we introduce multi-resolution convolution and deformable convolution operations. By enlarging the receptive field using convolution kernels with different dilation factors to capture temporal correlation information across different resolutions, and adaptively adjusting the sampling positions through additional offset vectors, we enhance the network's ability to capture correlated features between variables. Building upon this, we propose ATVCNet, an adaptive temporal-variable convolutional network designed to effectively model the local/global temporal dependencies and inter-variable dependencies of multivariate time series. Specifically, extracting and fusing time series features at different resolutions, captures both local contextual information and global patterns in the time series. The designed inter-variable feature adaptive extraction module captures the correlation among different variables in the time series. We evaluated the performance of ATVCNet across eight real-world datasets. The results indicate that ATVCNet achieved a performance improvement of approximately 63.4% over state-of-the-art MLSTF models.