arXiv subjects

Zhenshan Bing

Publications and source records attributed to Zhenshan Bing.

At least 37 records · Page 2Linked to original sources

arXiv2026

From Flow to One Step: Real-Time Multi-Modal Trajectory Policies via Implicit Maximum Likelihood Estimation-based Distribution Distillation

Generative policies based on diffusion and flow matching achieve strong performance in robotic manipulation by modeling multi-modal human demonstrations. However, their reliance on iterative Ordinary Differential Equation (ODE) integration introduces substantial latency, limiting high-frequency closed-loop control. Recent single-step acceleration methods alleviate this overhead but often exhibit distributional collapse, producing averaged trajectories that fail to execute coherent manipulation strategies. We propose a framework that distills a Conditional Flow Matching (CFM) expert into a fast single-step student via Implicit Maximum Likelihood Estimation (IMLE). A bi-directional Chamfer distance provides a set-level objective that promotes both mode coverage and fidelity, enabling preservation of the teacher multi-modal action distribution in a single forward pass. A unified perception encoder further integrates multi-view RGB, depth, point clouds, and proprioception into a geometry-aware representation. The resulting high-frequency control supports real-time receding-horizon re-planning and improved robustness under dynamic disturbances.

cs.RO↗

arXiv2026

M4Diffuser: Multi-View Diffusion Policy with Manipulability-Aware Control for Robust Mobile Manipulation

Mobile manipulation requires the coordinated control of a mobile base and a robotic arm while simultaneously perceiving both global scene context and fine-grained object details. Existing single-view approaches often fail in unstructured environments due to limited fields of view, exploration, and generalization abilities. Moreover, classical controllers, although stable, struggle with efficiency and manipulability near singularities. To address these challenges, we propose M4Diffuser, a hybrid framework that integrates a Multi-View Diffusion Policy with a novel Reduced and Manipulability-aware QP (ReM-QP) controller for mobile manipulation. The diffusion policy leverages proprioceptive states and complementary camera perspectives with both close-range object details and global scene context to generate task-relevant end-effector goals in the world frame. These high-level goals are then executed by the ReM-QP controller, which eliminates slack variables for computational efficiency and incorporates manipulability-aware preferences for robustness near singularities. Comprehensive experiments in simulation and real-world environments show that M4Diffuser achieves 7 to 56 percent higher success rates and reduces collisions by 3 to 31 percent over baselines. Our approach demonstrates robust performance for smooth whole-body coordination, and strong generalization to unseen tasks, paving the way for reliable mobile manipulation in unstructured environments. Details of the demo and supplemental material are available on our project website https://sites.google.com/view/m4diffuser.

cs.RO↗

arXiv2026

MOSAIC: Bridging the Sim-to-Real Gap in Generalist Humanoid Motion Tracking and Teleoperation with Rapid Residual Adaptation

Generalist humanoid motion trackers have recently achieved strong simulation metrics by scaling data and training, yet often remain brittle on hardware during sustained teleoperation due to interface- and dynamics-induced errors. We present MOSAIC, an open-source, full-stack system for humanoid motion tracking and whole-body teleoperation across multiple interfaces. MOSAIC first learns a teleoperation-oriented general motion tracker via RL on a multi-source motion bank with adaptive resampling and rewards that emphasize world-frame motion consistency, which is critical for mobile teleoperation. To bridge the sim-to-real interface gap without sacrificing generality, MOSAIC then performs rapid residual adaptation: an interface-specific policy is trained using minimal interface-specific data, and then distilled into the general tracker through an additive residual module, outperforming naive fine-tuning or continual learning. We validate MOSAIC with systematic ablations, out-of-distribution benchmarking, and real-robot experiments demonstrating robust offline motion replay and online long-horizon teleoperation under realistic latency and noise. Project page: baai-humanoid.github.io/MOSAIC.

cs.RO↗

arXiv2026

TacUMI: A Multi-Modal Universal Manipulation Interface for Contact-Rich Tasks

Task decomposition is critical for understanding and learning complex long-horizon manipulation tasks. Especially for tasks involving rich physical interactions, relying solely on visual observations and robot proprioceptive information often fails to reveal the underlying event transitions. This raises the requirement for efficient collection of high-quality multi-modal data as well as robust segmentation method to decompose demonstrations into meaningful modules. Building on the idea of the handheld demonstration device Universal Manipulation Interface (UMI), we introduce TacUMI, a multi-modal data collection system that integrates additionally ViTac sensors, force-torque sensor, and pose tracker into a compact, robot-compatible gripper design, which enables synchronized acquisition of all these modalities during human demonstrations. We then propose a multi-modal segmentation framework that leverages temporal models to detect semantically meaningful event boundaries in sequential manipulations. Evaluation on a challenging cable mounting task shows more than 90 percent segmentation accuracy and highlights a remarkable improvement with more modalities, which validates that TacUMI establishes a practical foundation for both scalable collection and segmentation of multi-modal demonstrations in contact-rich tasks.

cs.RO↗

arXiv2025

Learning effective pruning at initialization from iterative pruning

Pruning at initialization (PaI) reduces training costs by removing weights before training, which becomes increasingly crucial with the growing network size. However, current PaI methods still have a large accuracy gap with iterative pruning, especially at high sparsity levels. This raises an intriguing question: can we get inspiration from iterative pruning to improve the PaI performance? In the lottery ticket hypothesis, the iterative rewind pruning (IRP) finds subnetworks retroactively by rewinding the parameter to the original initialization in every pruning iteration, which means all the subnetworks are based on the initial state. Here, we hypothesise the surviving subnetworks are more important and bridge the initial feature and their surviving score as the PaI criterion. We employ an end-to-end neural network (\textbf{AutoS}parse) to learn this correlation, input the model's initial features, output their score and then prune the lowest score parameters before training. To validate the accuracy and generalization of our method, we performed PaI across various models. Results show that our approach outperforms existing methods in high-sparsity settings. Notably, as the underlying logic of model pruning is consistent in different models, only one-time IRP on one model is needed (e.g., once IRP on ResNet-18/CIFAR-10, AutoS can be generalized to VGG-16/CIFAR-10, ResNet-18/TinyImageNet, et al.). As the first neural network-based PaI method, we conduct extensive experiments to validate the factors influencing this approach. These results reveal the learning tendencies of neural networks and provide new insights into our understanding and research of PaI from a practical perspective. Our code is available at: https://github.com/ChengYaofeng/AutoSparse.git.

cs.CV↗

arXiv2025

OmniDexVLG: Learning Dexterous Grasp Generation from Vision Language Model-Guided Grasp Semantics, Taxonomy and Functional Affordance

Dexterous grasp generation aims to produce grasp poses that align with task requirements and human interpretable grasp semantics. However, achieving semantically controllable dexterous grasp synthesis remains highly challenging due to the lack of unified modeling of multiple semantic dimensions, including grasp taxonomy, contact semantics, and functional affordance. To address these limitations, we present OmniDexVLG, a multimodal, semantics aware grasp generation framework capable of producing structurally diverse and semantically coherent dexterous grasps under joint language and visual guidance. Our approach begins with OmniDexDataGen, a semantic rich dexterous grasp dataset generation pipeline that integrates grasp taxonomy guided configuration sampling, functional affordance contact point sampling, taxonomy aware differential force closure grasp sampling, and physics based optimization and validation, enabling systematic coverage of diverse grasp types. We further introduce OmniDexReasoner, a multimodal grasp type semantic reasoning module that leverages multi agent collaboration, retrieval augmented generation, and chain of thought reasoning to infer grasp related semantics and generate high quality annotations that align language instructions with task specific grasp intent. Building upon these components, we develop a unified Vision Language Grasping generation model that explicitly incorporates grasp taxonomy, contact structure, and functional affordance semantics, enabling fine grained control over grasp synthesis from natural language instructions. Extensive experiments in simulation and real world object grasping and ablation studies demonstrate that our method substantially outperforms state of the art approaches in terms of grasp diversity, contact semantic diversity, functional affordance diversity, and semantic consistency.

cs.RO↗

arXiv2025

DualGazeNet: A Biologically Inspired Dual-Gaze Query Network for Salient Object Detection

Recent salient object detection (SOD) methods aim to improve performance in four key directions: semantic enhancement, boundary refinement, auxiliary task supervision, and multi-modal fusion. In pursuit of continuous gains, these approaches have evolved toward increasingly sophisticated architectures with multi-stage pipelines, specialized fusion modules, edge-guided learning, and elaborate attention mechanisms. However, this complexity paradoxically introduces feature redundancy and cross-component interference that obscure salient cues, ultimately reaching performance bottlenecks. In contrast, human vision achieves efficient salient object identification without such architectural complexity. This contrast raises a fundamental question: can we design a biologically grounded yet architecturally simple SOD framework that dispenses with most of this engineering complexity, while achieving state-of-the-art accuracy, computational efficiency, and interpretability? In this work, we answer this question affirmatively by introducing DualGazeNet, a biologically inspired pure Transformer framework that models the dual biological principles of robust representation learning and magnocellular-parvocellular dual-pathway processing with cortical attention modulation in the human visual system. Extensive experiments on five RGB SOD benchmarks show that DualGazeNet consistently surpasses 25 state-of-the-art CNN- and Transformer-based methods. On average, DualGazeNet achieves about 60\% higher inference speed and 53.4\% fewer FLOPs than four Transformer-based baselines of similar capacity (VST++, MDSAM, Sam2unet, and BiRefNet). Moreover, DualGazeNet exhibits strong cross-domain generalization, achieving leading or highly competitive performance on camouflaged and underwater SOD benchmarks without relying on additional modalities.

cs.CV↗

arXiv2025

Inference-stage Adaptation-projection Strategy Adapts Diffusion Policy to Cross-manipulators Scenarios

Diffusion policies are powerful visuomotor models for robotic manipulation, yet they often fail to generalize to manipulators or end-effectors unseen during training and struggle to accommodate new task requirements at inference time. Addressing this typically requires costly data recollection and policy retraining for each new hardware or task configuration. To overcome this, we introduce an adaptation-projection strategy that enables a diffusion policy to perform zero-shot adaptation to novel manipulators and dynamic task settings, entirely at inference time and without any retraining. Our method first trains a diffusion policy in SE(3) space using demonstrations from a base manipulator. During online deployment, it projects the policy's generated trajectories to satisfy the kinematic and task-specific constraints imposed by the new hardware and objectives. Moreover, this projection dynamically adapts to physical differences (e.g., tool-center-point offsets, jaw widths) and task requirements (e.g., obstacle heights), ensuring robust and successful execution. We validate our approach on real-world pick-and-place, pushing, and pouring tasks across multiple manipulators, including the Franka Panda and Kuka iiwa 14, equipped with a diverse array of end-effectors like flexible grippers, Robotiq 2F/3F grippers, and various 3D-printed designs. Our results demonstrate consistently high success rates in these cross-manipulator scenarios, proving the effectiveness and practicality of our adaptation-projection strategy. The code will be released after peer review.

cs.RO↗

arXiv2025

Estimated Informed Anytime Search for Sampling-Based Planning via Adaptive Sampler

Path planning in robotics often involves solving continuously valued, high-dimensional problems. Popular informed approaches include graph-based searches, such as A*, and sampling-based methods, such as Informed RRT*, which utilize informed set and anytime strategies to expedite path optimization incrementally. Informed sampling-based planners define informed sets as subsets of the problem domain based on the current best solution cost. However, when no solution is found, these planners re-sample and explore the entire configuration space, which is time-consuming and computationally expensive. This article introduces Multi-Informed Trees (MIT*), a novel planner that constructs estimated informed sets based on prior admissible solution costs before finding the initial solution, thereby accelerating the initial convergence rate. Moreover, MIT* employs an adaptive sampler that dynamically adjusts the sampling strategy based on the exploration process. Furthermore, MIT* utilizes length-related adaptive sparse collision checks to guide lazy reverse search. These features enhance path cost efficiency and computation times while ensuring high success rates in confined scenarios. Through a series of simulations and real-world experiments, it is confirmed that MIT* outperforms existing single-query, sampling-based planners for problems in R^4 to R^16 and has been successfully applied to real-world robot manipulation tasks. A video showcasing our experimental results is available at: https://youtu.be/30RsBIdexTU

cs.RO↗

arXiv2025

Deep Fuzzy Optimization for Batch-Size and Nearest Neighbors in Optimal Robot Motion Planning

Efficient motion planning algorithms are essential in robotics. Optimizing essential parameters, such as batch size and nearest neighbor selection in sampling-based methods, can enhance performance in the planning process. However, existing approaches often lack environmental adaptability. Inspired by the method of the deep fuzzy neural networks, this work introduces Learning-based Informed Trees (LIT*), a sampling-based deep fuzzy learning-based planner that dynamically adjusts batch size and nearest neighbor parameters to obstacle distributions in the configuration spaces. By encoding both global and local ratios via valid and invalid states, LIT* differentiates between obstacle-sparse and obstacle-dense regions, leading to lower-cost paths and reduced computation time. Experimental results in high-dimensional spaces demonstrate that LIT* achieves faster convergence and improved solution quality. It outperforms state-of-the-art single-query, sampling-based planners in environments ranging from R^8 to R^14 and is successfully validated on a dual-arm robot manipulation task. A video showcasing our experimental results is available at: https://youtu.be/NrNs9zebWWk

cs.RO↗

arXiv2025

Language-Enhanced Mobile Manipulation for Efficient Object Search in Indoor Environments

Enabling robots to efficiently search for and identify objects in complex, unstructured environments is critical for diverse applications ranging from household assistance to industrial automation. However, traditional scene representations typically capture only static semantics and lack interpretable contextual reasoning, limiting their ability to guide object search in completely unfamiliar settings. To address this challenge, we propose a language-enhanced hierarchical navigation framework that tightly integrates semantic perception and spatial reasoning. Our method, Goal-Oriented Dynamically Heuristic-Guided Hierarchical Search (GODHS), leverages large language models (LLMs) to infer scene semantics and guide the search process through a multi-level decision hierarchy. Reliability in reasoning is achieved through the use of structured prompts and logical constraints applied at each stage of the hierarchy. For the specific challenges of mobile manipulation, we introduce a heuristic-based motion planner that combines polar angle sorting with distance prioritization to efficiently generate exploration paths. Comprehensive evaluations in Isaac Sim demonstrate the feasibility of our framework, showing that GODHS can locate target objects with higher search efficiency compared to conventional, non-semantic search strategies. Website and Video are available at: https://drapandiger.github.io/GODHS

cs.RO↗

arXiv2025

Genetic Informed Trees (GIT*): Path Planning via Reinforced Genetic Programming Heuristics

Optimal path planning involves finding a feasible state sequence between a start and a goal that optimizes an objective. This process relies on heuristic functions to guide the search direction. While a robust function can improve search efficiency and solution quality, current methods often overlook available environmental data and simplify the function structure due to the complexity of information relationships. This study introduces Genetic Informed Trees (GIT*), which improves upon Effort Informed Trees (EIT*) by integrating a wider array of environmental data, such as repulsive forces from obstacles and the dynamic importance of vertices, to refine heuristic functions for better guidance. Furthermore, we integrated reinforced genetic programming (RGP), which combines genetic programming with reward system feedback to mutate genotype-generative heuristic functions for GIT*. RGP leverages a multitude of data types, thereby improving computational efficiency and solution quality within a set timeframe. Comparative analyses demonstrate that GIT* surpasses existing single-query, sampling-based planners in problems ranging from R^4 to R^16 and was tested on a real-world mobile manipulation task. A video showcasing our experimental results is available at https://youtu.be/URjXbc_BiYg

cs.RO↗

arXiv2025

Elliptical K-Nearest Neighbors -- Path Optimization via Coulomb's Law and Invalid Vertices in C-space Obstacles

Path planning has long been an important and active research area in robotics. To address challenges in high-dimensional motion planning, this study introduces the Force Direction Informed Trees (FDIT*), a sampling-based planner designed to enhance speed and cost-effectiveness in pathfinding. FDIT* builds upon the state-of-the-art informed sampling planner, the Effort Informed Trees (EIT*), by capitalizing on often-overlooked information in invalid vertices. It incorporates principles of physical force, particularly Coulomb's law. This approach proposes the elliptical $k$-nearest neighbors search method, enabling fast convergence navigation and avoiding high solution cost or infeasible paths by exploring more problem-specific search-worthy areas. It demonstrates benefits in search efficiency and cost reduction, particularly in confined, high-dimensional environments. It can be viewed as an extension of nearest neighbors search techniques. Fusing invalid vertex data with physical dynamics facilitates force-direction-based search regions, resulting in an improved convergence rate to the optimum. FDIT* outperforms existing single-query, sampling-based planners on the tested problems in R^4 to R^16 and has been demonstrated on a real-world mobile manipulation task.

cs.RO↗

arXiv2025

Tree-Based Grafting Approach for Bidirectional Motion Planning with Local Subsets Optimization

Bidirectional motion planning often reduces planning time compared to its unidirectional counterparts. It requires connecting the forward and reverse search trees to form a continuous path. However, this process could fail and restart the asymmetric bidirectional search due to the limitations of lazy-reverse search. To address this challenge, we propose Greedy GuILD Grafting Trees (G3T*), a novel path planner that grafts invalid edge connections at both ends to re-establish tree-based connectivity, enabling rapid path convergence. G3T* employs a greedy approach using the minimum Lebesgue measure of guided incremental local densification (GuILD) subsets to optimize paths efficiently. Furthermore, G3T* dynamically adjusts the sampling distribution between the informed set and GuILD subsets based on historical and current cost improvements, ensuring asymptotic optimality. These features enhance the forward search's growth towards the reverse tree, achieving faster convergence and lower solution costs. Benchmark experiments across dimensions from R^2 to R^8 and real-world robotic evaluations demonstrate G3T*'s superior performance compared to existing single-query sampling-based planners. A video showcasing our experimental results is available at: https://youtu.be/3mfCRL5SQIU

cs.RO↗

arXiv2025

APT*: Asymptotically Optimal Motion Planning via Adaptively Prolated Elliptical R-Nearest Neighbors

Optimal path planning aims to determine a sequence of states from a start to a goal while accounting for planning objectives. Popular methods often integrate fixed batch sizes and neglect information on obstacles, which is not problem-specific. This study introduces Adaptively Prolated Trees (APT*), a novel sampling-based motion planner that extends based on Force Direction Informed Trees (FDIT*), integrating adaptive batch-sizing and elliptical $r$-nearest neighbor modules to dynamically modulate the path searching process based on environmental feedback. APT* adjusts batch sizes based on the hypervolume of the informed sets and considers vertices as electric charges that obey Coulomb's law to define virtual forces via neighbor samples, thereby refining the prolate nearest neighbor selection. These modules employ non-linear prolate methods to adaptively adjust the electric charges of vertices for force definition, thereby improving the convergence rate with lower solution costs. Comparative analyses show that APT* outperforms existing single-query sampling-based planners in dimensions from $\mathbb{R}^4$ to $\mathbb{R}^{16}$, and it was further validated through a real-world robot manipulation task. A video showcasing our experimental results is available at: https://youtu.be/gCcUr8LiEw4

cs.RO↗

arXiv2025

Direction Informed Trees (DIT*): Optimal Path Planning via Direction Filter and Direction Cost Heuristic

Optimal path planning requires finding a series of feasible states from the starting point to the goal to optimize objectives. Popular path planning algorithms, such as Effort Informed Trees (EIT*), employ effort heuristics to guide the search. Effective heuristics are accurate and computationally efficient, but achieving both can be challenging due to their conflicting nature. This paper proposes Direction Informed Trees (DIT*), a sampling-based planner that focuses on optimizing the search direction for each edge, resulting in goal bias during exploration. We define edges as generalized vectors and integrate similarity indexes to establish a directional filter that selects the nearest neighbors and estimates direction costs. The estimated direction cost heuristics are utilized in edge evaluation. This strategy allows the exploration to share directional information efficiently. DIT* convergence faster than existing single-query, sampling-based planners on tested problems in R^4 to R^16 and has been demonstrated in real-world environments with various planning tasks. A video showcasing our experimental results is available at: https://youtu.be/2SX6QT2NOek

cs.RO↗

arXiv2025

Whisker-based Active Tactile Perception for Contour Reconstruction

Perception using whisker-inspired tactile sensors currently faces a major challenge: the lack of active control in robots based on direct contact information from the whisker. To accurately reconstruct object contours, it is crucial for the whisker sensor to continuously follow and maintain an appropriate relative touch pose on the surface. This is especially important for localization based on tip contact, which has a low tolerance for sharp surfaces and must avoid slipping into tangential contact. In this paper, we first construct a magnetically transduced whisker sensor featuring a compact and robust suspension system composed of three flexible spiral arms. We develop a method that leverages a characterized whisker deflection profile to directly extract the tip contact position using gradient descent, with a Bayesian filter applied to reduce fluctuations. We then propose an active motion control policy to maintain the optimal relative pose of the whisker sensor against the object surface. A B-Spline curve is employed to predict the local surface curvature and determine the sensor orientation. Results demonstrate that our algorithm can effectively track objects and reconstruct contours with sub-millimeter accuracy. Finally, we validate the method in simulations and real-world experiments where a robot arm drives the whisker sensor to follow the surfaces of three different objects.

cs.RO↗

arXiv2025

Multi-Robot Assembly of Deformable Linear Objects Using Multi-Modal Perception

Industrial assembly of deformable linear objects (DLOs) such as cables offers great potential for many industries. However, DLOs pose several challenges for robot-based automation due to the inherent complexity of deformation and, consequentially, the difficulties in anticipating the behavior of DLOs in dynamic situations. Although existing studies have addressed isolated subproblems like shape tracking, grasping, and shape control, there has been limited exploration of integrated workflows that combine these individual processes. To address this gap, we propose an object-centric perception and planning framework to achieve a comprehensive DLO assembly process throughout the industrial value chain. The framework utilizes visual and tactile information to track the DLO's shape as well as contact state across different stages, which facilitates effective planning of robot actions. Our approach encompasses robot-based bin picking of DLOs from cluttered environments, followed by a coordinated handover to two additional robots that mount the DLOs onto designated fixtures. Real-world experiments employing a setup with multiple robots demonstrate the effectiveness of the approach and its relevance to industrial scenarios.

cs.RO↗