Papers

@conference{li2025learning,
title = {Learning to Optimize Package Picking for Large-Scale, Real-World Robot Induction},
author = {S Li and A Keipour and S Zhao and S Rajagopalan and C Swan and K Bekris},
url = {https://arxiv.org/abs/2506.09765},
year = {2025},
date = {2025-07-06},
booktitle = {19th International Symposium on Experimental Robotics (ISER)},
abstract = {Warehouse automation plays a pivotal role in enhancing operational efficiency, minimizing costs, and improving resilience to workforce variability. While prior research has demonstrated the potential of machine learning (ML) models to increase picking success rates in large-scale robotic fleets by prioritizing high-probability picks and packages, these efforts primarily focused on predicting success probabilities for picks sampled using heuristic methods. Limited attention has been given, however, to leveraging data-driven approaches to directly optimize sampled picks for better performance at scale. In this study, we propose an ML-based framework that predicts transform adjustments as well as improving the selection of suction cups for multi-suction end effectors for sampled picks to enhance their success probabilities. The framework was integrated and evaluated in test workcells that resemble the operations of Amazon Robotics' Robot Induction (Robin) fleet, which is used for package manipulation. Evaluated on over 2 million picks, the proposed method achieves a 20\% reduction in pick failure rates compared to a heuristic-based pick sampling baseline, demonstrating its effectiveness in large-scale warehouse automation scenarios. },
keywords = {Learning, Manipulation},
pubstate = {published},
tppubtype = {conference}
}

@conference{wang2025Demonstrating,
title = {Demonstrating Item Picking at Scale via Effective Learning of Multimodal Representations},
author = {C Wang and J Vanbaar and C Mitash and S Li and D Randle and A Keipour and M Hussein and K Bekris and K Katyal},
url = {https://arxiv.org/abs/2506.10359},
year = {2025},
date = {2025-06-21},
booktitle = {Proceedings of Robotics: Science and Systems (RSS)},
journal = {Proceedings of Robotics: Science and Systems (RSS)},
abstract = {This work demonstrates how autonomously learning aspects of robotic operation from sparsely-labeled, real-world data of deployed, engineered solutions at industrial scale can provide with solutions that achieve improved performance. Specifically, it focuses on multi-suction robot picking and performs a comprehensive study on the application of multi-modal visual encoders for predicting the success of candidate robotic picks. Picking diverse items from unstructured piles is an important and challenging task for robot manipulation in real-world settings, such as warehouses. Methods for picking from clutter must work for an open set of items while simultaneously meeting latency constraints to achieve high throughput. The demonstrated approach utilizes multiple input modalities, such as RGB, depth and semantic segmentation, to estimate the quality of candidate multi-suction picks. The strategy is trained from real-world item picking data, with a combination of multimodal pretrain and finetune. The manuscript provides comprehensive experimental evaluation performed over a large item-picking dataset, an item-picking dataset targeted to include partial occlusions, and a package-picking dataset, which focuses on containers, such as boxes and envelopes, instead of unpackaged items. The evaluation measures performance for different item configurations, pick scenes, and object types. Ablations help to understand the effects of in-domain pretraining, the impact of different modalities and the importance of finetuning. These ablations reveal both the importance of training over multiple modalities but also the ability of models to learn during pretraining the relationship between modalities so that during finetuning and inference, only a subset of them can be used as input. },
keywords = {Learning, Manipulation, Planning},
pubstate = {published},
tppubtype = {conference}
}

@conference{Wen:2020ab,
title = {se(3)-TrackNet: Data-Driven 6d Pose Tracking by Calibrating Image Residuals in Synthetic Domains},
author = {B Wen and C Mitash and B Ren and K Bekris},
url = {http://arxiv.org/abs/2007.13866},
year = {2020},
date = {2020-10-01},
booktitle = {IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
address = {Las Vegas, NV},
abstract = {Tracking the 6D pose of objects in video sequences is important for robot manipulation. This task, however, introduces multiple challenges: (i) robot manipulation involves significant occlusions; (ii) data and annotations are troublesome and difficult to collect for 6D poses, which complicates machine learning solutions, and (iii) incremental error drift often accumulates in long term tracking to necessitate re-initialization of the object's pose. This work proposes a data-driven optimization approach for long-term, 6D pose tracking. It aims to identify the optimal relative pose given the current RGB-D observation and a synthetic image conditioned on the previous best estimate and the object's model. The key contribution in this context is a novel neural network architecture, which appropriately disentangles the feature encoding to help reduce domain shift, and an effective 3D orientation representation via Lie Algebra. Consequently, even when the network is trained only with synthetic data can work effectively over real images. Comprehensive experiments over benchmarks - existing ones as well as a new dataset with significant occlusions related to object manipulation - show that the proposed approach achieves consistently robust estimates and outperforms alternatives, even though they have been trained with real images. The approach is also the most computationally efficient among the alternatives and achieves a tracking frequency of 90.9Hz.},
keywords = {Learning, Perception},
pubstate = {published},
tppubtype = {conference}
}

@conference{Mitash:2019aa,
title = {Scene-Level Pose Estimation for Multiple Instances of Densely Packed Objects},
author = {C Mitash and B Wen and K Bekris and A Boularias},
url = {https://arxiv.org/pdf/1910.04953.pdf},
year = {2019},
date = {2019-10-01},
booktitle = {Conference on Robot Learning (CoRL)},
address = {Osaka, Japan},
abstract = {This paper introduces key machine learning operations that allow the realization of robust, joint 6D pose estimation of multiple instances of objects either densely packed or in unstructured piles from RGB-D data. The first objective is to learn semantic and instance-boundary detectors without manual labeling. An adversarial training framework in conjunction with physics-based simulation is used to achieve detectors that behave similarly in synthetic and real data. Given the stochastic output of such detectors, candidates for object poses are sampled. The second objective is to automatically learn a single score for each pose candidate that represents its quality in terms of explaining the entire scene via a gradient boosted tree. The proposed method uses features derived from surface and boundary alignment between the observed scene and the object model placed at hypothesized poses. Scene-level, multi-instance pose estimation is then achieved by an integer linear programming process that selects hypotheses that maximize the sum of the learned individual scores, while respecting constraints, such as avoiding collisions. To evaluate this method, a dataset of densely packed objects with challenging setups for state-of-the-art approaches is collected. Experiments on this dataset and a public one show that the method significantly outperforms alternatives in terms of 6D pose accuracy while trained only with synthetic datasets.},
keywords = {Learning, Perception},
pubstate = {published},
tppubtype = {conference}
}