Dual-arm robotic grasping is essential for manipulating large, heavy, and geometrically complex objects that cannot be reliably handled using a single manipulator. These large objects often contain only sparse graspable regions determined by local geometric properties such as thickness, edge structure, and gripper clearance. Prior bimanual grasping methods assume access to a full point cloud of the object which inherently contains this geometric information, but may not be accessible in real scenarios. This work proposes PartialBiGrasp, a dual-arm grasp generation framework that operates directly on partial point cloud observations. Our model learns geometric features implicitly through convolutional occupancy networks, enabling local reasoning about graspability, collision-free contact regions, and object thickness. We leverage this understanding to generate force-closure compliant grasp pairs, which are further refined using a sampling-based optimization to correct for ambiguity caused by incomplete geometry. We evaluate our approach using analytical force-closure metrics, large-scale simulation experiments, and real-world robot evaluations on noisy partial point clouds of novel objects, demonstrating robust and physically stable dual-arm grasp generation.
Figures & tables
Figure 1: Partial observations hide grasp-relevant geometry and introduce deceptive grasp cues, often leading to unstable grasps, as denoted by the red grasps. PartialBiGrasp overcomes these challenges, shown as the green grasps, by reasoning about hidden local geometry to generate physically feasible dual-arm grasps.
Figure 2: Given a partial point cloud P , the Global Encoder produces a continuous object representation from which the Grasp Generation Module G decodes single-arm grasps. High-graspability candidates are paired and scored by the Force Closure Module FC . Local geometry is captured by encoding gripper-frame point cloud crops PG with the Local Encoder. Finally, a sampling-based optimizer uses local occupancy predictions occl to reduce collisions and enforce stable contact, yielding force-closure-stable grasps that are aware of hidden local geometry.
Figure 3: Local Occupancy Guided Refinement implicitly identifies unknown properties such as local thickness and surface continuation under occlusion, and successfully updates grasp poses in local regions, maintaining force closure stability while reducing collisions. Red signifies the initial pose, and Green signifies the refined pose.
Figure 4: Qualitative comparison. Existing methods frequently suffer from occlusion artifacts, reconstruction errors, and unreliable grasp pairing, leading to poorly aligned or unstable dual-arm grasps. Our framework combines global and local geometric reasoning to generate well-aligned, collision-free, and force-closure-stable grasp pairs. Red circles correspond to collisions due to pose error, and coloured squares are the VLM bounding boxes.
Method
DG16M
Real World Dataset
FC% ↑
GS% ↑
PC% ↓
Cov. ↑
FC% ↑
GS% ↑
PC% ↓
ContactGraspNet + Attn Pairing
21.10
52.14
47.41
10.11
22.56
65.04
57.15
ContactGraspNet + Classifier Pairing
15.03
50.05
63.44
9.42
22.22
62.60
63.44
DAGDiff (Partial)
22.35
46.63
49.19
14.48
17.47
62.47
59.27
Recgen + DAGDiff (Full)
–
–
–
–
20.51
64.33
53.2
Single GraspGen + VLM
–
–
–
–
22.96
66.93
50.16
Table 1: Quantitative comparison of grasp generation methods on DG16M and real-world dataset.
Method
DG16M
Real World Dataset
FC% ↑
GS% ↑
PC% ↓
Cov. ↑
FC% ↑
GS% ↑
PC% ↓
No Refinement
45.35
61.51
32.61
15.09
40.74
76.47
44.68
Global Refinement
39.57
63.67
35.30
14.42
40.85
74.40
42.74
No Residual
19.85
41.70
67.63
7.15
27.02
65.31
48.50
Ours
55.16
67.87
17.48
16.96
51.06
81.54
38.80
Table 2: Ablation study of different components of our method on DG16M and the real-world dataset.
Figure 5: Real-World Demos. For the objects in the Real World testing set, we also execute grasps on a heterogenous dual-arm setup consisting of an xArm7 and xArm6 Lite. Note: The briefcase has no grasps on the sides due to it being thicker than the average gripper width.
Appendix figures & tables11 assets
Supplementary material from the paper’s appendix.
Appendix
Figure 6: Graspability map for a vase. Red points indicate high-graspability regions, while gray points indicate lower-graspability regions.
Figure 9
Figure 9: Triplane representations extracted from the input image. For both (A) Global and (B) Local Triplanes, the left panel shows the initial triplane features and the right panel shows the features after U-Net processing.
Figure 10: Coarse global occupancy output from the Global Encoder.
Figure 11: Qualitative examples of occupancy-guided grasp refinement. The original network prediction is shown in red, while the final refined grasp is shown in green. Refinement improves alignment with the object geometry and yields more stable grasp configurations.
Figure 12: Control points used for occupancy-guided grasp refinement. Outer control points are used to evaluate collisions with the object, while inner control points are used to assess contact quality.
wi=∑jexp(−λJ(ξj)−Jmin)exp(−λJ(ξi)−Jmin)
Appendix
Algorithm 1 Occupancy-Guided Grasp Refinement
Figure 13: Contact-GraspNet baseline for dual-arm grasp generation. Independent single-arm grasp candidates are generated from a partial point cloud and subsequently paired using either an attention-based transformer architecture or a classifier-based pairing network to identify compatible grasp pairs.
Figure 14: Real-world grasp generation using RecGen and DAGDiff. A complete object mesh is reconstructed from a single RGB-D view using RecGen, and DAGDiff generates dual-arm grasp pairs from points sampled on the reconstructed mesh.
Figure 15: VLM-guided grasp pairing baseline. A vision-language model predicts compatible grasping regions from an image, while a single-arm grasp generator produces grasp candidates from the partial point cloud. Dual-arm grasp pairs are obtained by matching projected grasp contacts to the predicted regions and selecting the top-ranked pairings.
Figure 16: Real-world evaluation setup consisting of an xArm7 and an xArm6 Lite equipped with parallel-jaw grippers. A RealSense D455 RGB-D camera captures single-view observations of the workspace, and an ArUco board is used for camera extrinsics.
Bimanual dexterous grasping of large objects is a critical challenge in robotic manipulation. However, most existing studies focus on sequential manipulation rather than cooperative grasping, and methods addressing such bimanual tasks have largely been limited to simulation. These limitations stem from the difficulty of acquiring full 3D object models and generating physically plausible grasping actions. To fill this gap, we propose a real-world bimanual grasping framework that includes: a multimodal dataset capturing joint angles, visual observations and force signals; a Denoising Diffusion Probabilistic Model (DDPM)-based module that generates joint-level grasp configurations from segmented point clouds; and an execution strategy that integrates motion planning with online grasp refinement to ensure physical stability and feasibility. Our approach enables the synthesis of executable bimanual grasps from single-view inputs, reducing dependence on complete 3D object models and ensuring stable real-world performance. Experiments on a dual-arm robot demonstrate high success rates across unseen objects with varying geometries and poses, and ablation studies confirm the contributions of key components of our system.
Ziming Li, Mingxuan Wu, Jiaqi Zhang +4
School of Computer Science, The University of Auckland, Auckland 1010, New Zealand · College of Computer Science, Chongqing University, Chongqing 401331, China · College of Electronic and Information Engineering, Southwest University, Chongqing 400715, China
Robotic grasping from single-view observations remains a critical challenge in manipulation. However, existing methods still struggle to generate reliable grasp candidates and stably evaluate grasp feasibility under incomplete geometric information. To address these limitations, we present SuperGrasp, a new two-stage framework for single-view parallel-jaw grasping. In the first stage, we introduce a Similarity Matching Module that efficiently retrieves valid and diverse grasp candidates by matching the input single-view point cloud with a precomputed primitive dataset based on superquadric coefficients. In the second stage, we propose E-RNet, an end-to-end network that expands the grasp-aware region and takes the initial grasp closure region as a local anchor region, capturing the contextual relationship between the local region and its surrounding spatial neighborhood, thereby enabling more accurate and reliable grasp evaluation and introducing small-range local refinement to improve grasp adaptability. To enhance generalization, we construct a primitive dataset containing 1.2k standard geometric primitives for similarity matching and collect a point cloud dataset of 100k samples from 124 objects, annotated with stable grasp labels for network training. Extensive experiments in both simulation and real-world environments demonstrate that our method achieves stable grasping performance and good generalization across novel objects and clutter scenes.
Bimanual dexterous grasping is a fundamental and promising area in robotics, yet its progress is constrained by the lack of comprehensive datasets and powerful generation models. In this work, we propose BiDexGrasp, consisting of a large-scale bimanual dexterous grasp dataset and a novel learning-based framework. For dataset construction, we propose a novel bimanual grasp synthesis pipeline to efficiently annotate physically feasible data. This pipeline addresses the challenges of high-dimensional bimanual grasping through a two-stage synthesis strategy of efficient region-based grasp initialization and decoupled force-closure grasp optimization. Powered by this pipeline, we construct a large-scale bimanual dexterous grasp dataset, comprising 6351 diverse objects with sizes ranging from 30 to 80 cm, along with 9.53 million annotated grasp data. Based on this dataset, we further introduce a novel learning-based dexterous grasping generation framework. The framework lies in two key designs: a bimanual coordination module and a geometry-size-adaptive grasp generation strategy to generate coordinated and high-quality grasps on unseen objects. Extensive experiments conducted in both simulation and real world demonstrate the superior performance of our proposed data synthesis pipeline and learned generative framework.
Mu Lin, Yi-Lin Wei, Jiaxuan Chen +7
Sun Yat-sen University · Peking University · Tsinghua University