A Unified Multi-Task Deep Learning Framework for Robotic Bin-Picking of Planar Objects

Towards a Modular Bin-picking Framework for Handling Object Pose Uncertainties

In recent years, there has been growing interest in robust robotic systems for precise bin-picking applications. To achieve reliable performance, such systems must address errors arising from both the object pose estimation and the grasping process. Although various approaches have been proposed, they typically target specific challenges and do not offer general solutions. In this paper, we present a modular framework that jointly handles both error types. The framework incorporates object pose distribution estimation to account for pose uncertainty, which frequently arises in situations with ambiguous observations where a single correct pose cannot be determined. To further reduce uncertainty, we introduce a second-viewpoint module that computes complementary pose distributions, which are subsequently fused. This fusion decreases overall uncertainty and improves system efficiency. Additionally, two independent modules are included to compensate for grasping errors. The modular design allows the components to be combined for optimal performance or used individually, depending on the physical setup. The proposed method is evaluated in a real-world setup with three different objects, with no errors, and all modules are shown to improve efficiency. These results suggest that incorporating pose distributions with grasping pose errors is a promising direction for developing more flexible and reliable robotic production systems. To the best of our knowledge, this is the first framework that jointly addresses both grasping and object pose uncertainties using interchangeable modules. We believe there is ample opportunity to integrate additional modules, resulting in improved performance and flexibility. The current framework is limited to pose uncertainties in SO(2), but it could be extended to SE(3), enabling additional modules to improve the system.

Model-agnostic pose estimation for enhanced collaborative robot grasping via binocular vision

A novel 7-DoF grasping pose generation framework that integrates sparse attention and null convolution is introduced, which enhances the model’s ability to capture fine-grained features from point clouds, significantly improving the accuracy of parallel gripping pose estimation.

Conditional Flow Matching for Grasp Pose Generation in Multi-Workpiece Scenes

While most robotic research focuses on household tasks such as bus table arrangement and cloth folding, numerous manipulation tasks remain challenging for industrial applications, particularly the grasping and transportation of scattered workpieces. In this paper, we propose SGDIFF, a vision-guided grasping architecture that integrates a pre-trained vision-language model (VLM) with flow matching-based diffusion model. Given an RGB-D image of a tabletop scene, the VLM first detects each workpiece, outputs its bounding box, and assigns a unique ID. A point cloud is then generated for each detected instance. Subsequently, flow matching model iteratively refines the initially noisy gripper pose to a stable and collision-free grasp for each target workpiece. The proposed method eliminates the need for object-specific models and enables efficient multi-object grasping in cluttered industrial environments. The experiments validates the effectiveness of combining semantic understanding with precise pose refinement for robust industrial automation.

FreeTrack6d: a training-free framework for 6D pose tracking and robotic grasping of moving objects on conveyor belts

For robotic dynamic grasping of moving objects in conveyor-belt scenarios, accurate and robust 6D pose estimation and tracking are essential for reliable grasping. However, existing deep-learning-based methods usually rely on large amounts of supervised data for specific objects or categories, which limits their generalization, deployment efficiency, and flexibility for rapid object changeover in industrial applications. To address these challenges, this paper proposes FreeTrack6D, a training-free unified segmentation and 6D pose tracking framework. Relying only on the CAD model of the target object, FreeTrack6D can be applied to dynamic tracking and grasping of unseen objects without additional object-specific training. Specifically, an adaptive multi-cue mask generation module is first introduced to generate frame-wise target masks in real time, which provides target-region constraints for initial pose registration and subsequent pose refinement. This helps reduce the influence of background interference and target-region misalignment caused by rapid motion. Based on the generated mask, RGB-D observations, and the CAD model, FoundationPose is used for initial 6D pose registration and subsequent pose refinement. To improve tracking robustness under large inter-frame motion and rotational variations, a Kalman-guided multi-hypothesis refinement strategy is further designed, where multiple candidate poses predicted from historical motion states are refined and selected according to mask consistency. In addition, a Pose Consistency-aware Association and Gating mechanism is developed to reject abnormal detections, protect the filter state, and trigger re-initialization when consecutive mismatches occur. By integrating frame-wise mask generation, multi-hypothesis pose refinement, motion prediction, observation gating, and visual-servo-based robot control, FreeTrack6D forms a closed-loop training-free dynamic grasping framework.

Iterative Grasp Pose Refinement: A Deep Reinforcement Learning Approach for 2D Vision

A reinforcement learning-based framework for robotic grasp refinement, integrating keypoint-based object representations with a Deep Q-Network (DQN), is proposed, offering a scalable and adaptable solution for contact-rich manipulation tasks.

An improved, scalable and modular framework for vision-based robotic grasp detection