Grasp learning in noisy environments, such as occlusions, sensor noise, and out-of-distribution (OOD) objects, poses significant challenges. Recent learning-based approaches focus primarily on capturing aleatoric uncertainty from inherent data noise. The epistemic uncertainty, which represents the OOD recognition, is often addressed by ensembles with multiple forward paths, limiting real-time application. In this paper, we propose an uncertainty-aware approach for 6-DoF grasp detection using evidential learning to comprehensively capture both uncertainties in real-world robotic grasping. As a key contribution, we introduce vMF-Contact, a novel architecture for learning hierarchical contact grasp representations with probabilistic modeling of directional uncertainty as von Mises-Fisher (vMF) distribution. To achieve this, we analyze the theoretical formulation of the second-order objective on the posterior parametrization, providing formal guarantees for the model's ability to quantify uncertainty and improve grasp prediction performance. Moreover, we enhance feature expressiveness by applying partial point reconstructions as an auxiliary task, improving the comprehension of uncertainty quantification as well as the generalization to unseen objects. In the real-world experiments, our method demonstrates a significant improvement by 39% in the overall clearance rate compared to the baselines. The code is available under: https://github.com/YitianShi/vMF-Contact/tree/main

Grasping unknown objects in unstructured environments is one of the most challenging and demanding tasks for robotic bin picking systems. Developing a holistic approach is crucial to building such dexterous bin picking systems to meet practical requirements on speed, cost and reliability. Proposed datasets so far focus only on challenging sub-problems and are therefore limited in their ability to leverage the complementary relationship between individual tasks. In this paper, we tackle this holistic data challenge and design MetaGraspNetV2, an all-in-one bin picking dataset consisting of (i) a photo-realistic dataset with over 296k images, which has been created through physics-based metaverse synthesis; and (ii) a real-world test dataset with 3.2k images featuring task-specific difficulty levels. Both datasets provide full annotations for amodal panoptic segmentation, object relationship detection, occlusion reasoning, 6-DoF pose estimation, and grasp detection for a parallel-jaw as well as a vacuum gripper. Extensive experiments demonstrate that our dataset outperforms state-of-the-art datasets in object detection, instance segmentation, amodal detection, parallel-jaw grasping, and vacuum grasping. Furthermore, leveraging the potential of our data for building holistic perception systems, we propose a single-shot-multi-pick (SSMP) grasping policy for scene understanding accelerated fast picking in high clutter. SSMP reasons about suitable manipulation orders for blindly picking multiple items given a single image acquisition. Physical robot experiments demonstrate that SSMP effectively speeds up cycle times through reducing image acquisitions by more than 47% while providing better grasp performance compared to state-of-the-art bin picking methods.Note to Practitioners—In robotic bin picking, most proposed methods and datasets focus on solving only one aspect of the grasping task, such as grasp point detection, object detection, or relationship reasoning. They do not address practical aspects such as the widespread use of vacuum grasp technology or the need for short cycle times. In practice, however, efficient bin picking solutions often rely on multiple task-specific methods. Hence, having one dataset for a large variety of vision-related tasks in robotic picking reduces data redundancy and enables the development of holistic methods. While deep learning has been proven highly effective for bin picking vision systems, it demands large, high-quality training datasets. Collecting such datasets in the real-world, while assuring label quality and consistency, is prohibitively expensive and time-consuming. To overcome these challenges, we set up a photo-realistic metaverse data generation pipeline and create a large-scale synthetic training dataset. Furthermore, we design a comprehensive real-world dataset for testing. Unlike previously proposed datasets, our datasets provide difficulty levels and annotations in simulation and real-world for a comprehensive list of high-level tasks, including amodal object detection, scene layout reasoning, and grasp detection. In real-world applications, cycle time is a critical factor affecting the productivity and profitability of a robotic system. We tackle time-efficiency through scene understanding and demonstrate the capability of our data regarding holistic system development by proposing a single-shot-multi-pick (SSMP) policy. Our SSMP algorithm, trained exclusively on our synthetic data, distinguishes between uncovered and occluded items, and infers specific manipulation orders to perform multiple blind picks in a single shot. Physical robot experiments show that SSMP was able to reduce image acquisitions by more than 47% without compromising grasp performance. This clearly demonstrates that SSMP, together with our dataset, paves the way for application-oriented research in time-critical bin picking.

Abstract Knowing the robot's pose is a crucial prerequisite for mobile robot tasks such as collision avoidance or autonomous navigation. Using powerful predictive models to estimate transformations for visual odometry via downward facing cameras is an understudied area of research. This work proposes a novel approach based on deep learning for estimating ego motion with a downward looking camera. The network can be trained completely unsupervised and is not restricted to a specific motion model. We propose two neural network architectures based on the Early Fusion and Slow Fusion design principle: “EarlyBird” and “SlowBird”. Both networks share a Spatial Transformer layer for image warping and are trained with a modified structural similarity index (SSIM) loss function. Experiments carried out in simulation and for a real world differential drive robot show similar and partially better results of our proposed deep learning based approaches compared to a state-of-the-art method based on fast Fourier transformation.