Search

Domain randomization in reinforcement learning is an established technique for increasing the robustness of control policies trained in simulation. By randomizing environment properties during training, the learned policy can become robust to uncertainties along the randomized dimensions. While the environment distribution is typically specified by hand, in this paper we investigate automatically discovering a sampling distribution via entropy-regularized reward maximization of a normalizing-flow-based neural sampling distribution. We show that this architecture is more flexible and provides greater robustness than existing approaches that learn simpler, parameterized sampling distributions, as demonstrated in six simulated and one real-world robotics domain. Lastly, we explore how these learned sampling distributions, combined with a privileged value function, can be used for out-of-distribution detection in an uncertainty-aware multi-step manipulation planner.

The scarcity of labeled action data poses a considerable challenge for developing machine learning algorithms for robotic object manipulation. It is expensive and often infeasible for a robot to interact with many objects. Conversely, visual data of objects, without interaction, is abundantly available and can be leveraged for pretraining and feature extraction. However, current methods that rely on image data for pretraining do not easily adapt to task-specific predictions, since the learned features are not guaranteed to be relevant. This paper introduces the Semi-Supervised Neural Process (SSNP): an adaptive reward-prediction model designed for scenarios in which only a small subset of objects have labeled interaction data. In addition to predicting reward labels, the latent-space of the SSNP is jointly trained with an autoencoding objective using passive data from a much larger set of objects. Jointly training with both types of data allows the model to focus more effectively on generalizable features and minimizes the need for extensive retraining, thereby reducing computational demands. The efficacy of SSNP is demonstrated through a door-opening task, leading to better performance than other semi-supervised methods, and only using a fraction of the data compared to other adaptive models.

We present FORGE, a method for sim-to-real transfer of force-aware manipulation policies in the presence of significant pose uncertainty. During simulation-based policy learning, FORGE combines a force threshold mechanism with a dynamics randomization scheme to enable robust transfer of the learned policies to the real robot. At deployment, FORGE policies, conditioned on a maximum allowable force, adaptively perform contact-rich tasks while avoiding aggressive and unsafe behaviour, regardless of the controller gains. Additionally, FORGE policies predict task success, enabling efficient termination and autonomous tuning of the force threshold. We show that FORGE can be used to learn a variety of robust contact-rich policies, including the forceful insertion of snap-fit connectors. We further demonstrate the multistage assembly of a planetary gear system, which requires success across three assembly tasks: nut threading, insertion, and gear meshing. Project website can be accessed at https://noseworm.github.io/forge/., Comment: IndustReal comparisons and snap-fit task added (v2)

We selected 29 medical imaging projects from 48 candidates, assessed 10 software qualities by answering 108 questions for each software project, and interviewed 8 of the 29 development teams. Based on the quantitative data, we ranked the MI software with the Analytic Hierarchy Process (AHP). The four top-ranked software products are 3D Slicer, ImageJ, Fiji, and OHIF Viewer. Generally, MI software is in a healthy state as shown by the following: we observed 88% of the documentation artifacts recommended by research software development guidelines, 100% of MI projects use version control tools, and developers appear to use the common quasi-agile research software development process. However, the current state of the practice deviates from the existing guidelines because of the rarity of some recommended artifacts, low usage of continuous integration (17% of the projects), low use of unit testing (about 50% of projects), and room for improvement with documentation (six of nine developers felt their documentation was not clear enough). From interviewing the developers, we identified five pain points and two qualities of potential concern: lack of development time, lack of funding, technology hurdles, ensuring correctness, usability, maintainability, and reproducibility. The interviewees proposed strategies to improve the state of the practice, to address the identified pain points, and to improve software quality. Combining their ideas with ours, we have the following list of recommendations: increase documentation, increase testing by enriching datasets, increase continuous integration usage, move to web applications, employ linters, use peer reviews, design for change, add assurance cases, and incorporate a "Generate All Things" approach., Comment: 73 pages, 14 figures, 12 tables

Robots cannot yet match humans' ability to rapidly learn the shapes of novel 3D objects and recognize them robustly despite clutter and occlusion. We present Bayes3D, an uncertainty-aware perception system for structured 3D scenes, that reports accurate posterior uncertainty over 3D object shape, pose, and scene composition in the presence of clutter and occlusion. Bayes3D delivers these capabilities via a novel hierarchical Bayesian model for 3D scenes and a GPU-accelerated coarse-to-fine sequential Monte Carlo algorithm. Quantitative experiments show that Bayes3D can learn 3D models of novel objects from just a handful of views, recognizing them more robustly and with orders of magnitude less training data than neural baselines, and tracking 3D objects faster than real time on a single GPU. We also demonstrate that Bayes3D learns complex 3D object models and accurately infers 3D scene composition when used on a Panda robot in a tabletop scenario.

In this paper, we investigate a scenario in which a robot learns a low-dimensional representation of a door given a video of the door opening or closing. This representation can be used to infer door-related parameters and predict the outcomes of interacting with the door. Current machine learning based approaches in the doors domain are based primarily on labelled datasets. However, the large quantity of available door data suggests the feasibility of a semisupervised approach based on pretraining. To exploit the hierarchical structure of the dataset where each door has multiple associated images, we pretrain with a structured latent variable model known as a neural statistician. The neural satsitician enforces separation between shared context-level variables (common across all images associated with the same door) and instance-level variables (unique to each individual image). We first demonstrate that the neural statistician is able to learn an embedding that enables reconstruction and sampling of realistic door images. Then, we evaluate the correspondence of the learned embeddings to human-interpretable parameters in a series of supervised inference tasks. It was found that a pretrained neural statistician encoder outperformed analogous context-free baselines when predicting door handedness, size, angle location, and configuration from door images. Finally, in a visual bandit door-opening task with a variety of door configuration, we found that neural statistician embeddings achieve lower regret than context-free baselines.

We present Queer in AI as a case study for community-led participatory design in AI. We examine how participatory design and intersectional tenets started and shaped this community's programs over the years. We discuss different challenges that emerged in the process, look at ways this organization has fallen short of operationalizing participatory and intersectional principles, and then assess the organization's impact. Queer in AI provides important lessons and insights for practitioners and theorists of participatory methods broadly through its rejection of hierarchy in favor of decentralization, success at building aid and programs by and for the queer community, and effort to change actors and institutions outside of the queer community. Finally, we theorize how communities like Queer in AI contribute to the participatory design in AI more broadly by fostering cultures of participation in AI, welcoming and empowering marginalized participants, critiquing poor or exploitative participatory practices, and bringing participation to institutions outside of individual research projects. Queer in AI's work serves as a case study of grassroots activism and participatory methods within AI, demonstrating the potential of community-led participatory methods and intersectional praxis, while also providing challenges, case studies, and nuanced insights to researchers developing and using participatory methods., Comment: To appear at FAccT 2023

The rapid integration of artificial intelligence across traditional research domains has generated an amalgamation of nomenclature. As cross-discipline teams work together on complex machine learning challenges, finding a consensus of basic definitions in the literature is a more fundamental problem. As a step in the Delphi process to define issues with trust and barriers to the adoption of autonomous systems, our study first collected and ranked the top concerns from a panel of international experts from the fields of engineering, computer science, medicine, aerospace, and defence, with experience working with artificial intelligence. This document presents a summary of the literature definitions for nomenclature derived from expert feedback., Comment: 6 pages

Researchers have increasingly turned to crowdfunding platforms to gain insights into entrepreneurial activity and dynamics. While previous studies have explored various factors influencing crowdfunding success, such as technology, communication, and marketing strategies, the role of visual elements that can be automatically extracted from images has received less attention. This is surprising, considering that crowdfunding platforms emphasize the importance of attention-grabbing and high-resolution images, and previous research has shown that image characteristics can significantly impact product evaluations. Indeed, a comprehensive review of empirical articles (n = 202) that utilized Kickstarter data, focusing on the incorporation of visual information in their analyses. Our findings reveal that only 29.70% controlled for the number of images, and less than 12% considered any image details. In this manuscript, we review the literature on image processing and its relevance to the business domain, highlighting two types of visual variables: visual counts (number of pictures and number of videos) and image details. Building upon previous work that discussed the role of color, composition and figure-ground relationships, we introduce visual scene elements that have not yet been explored in crowdfunding, including the number of faces, the number of concepts depicted, and the ease of identifying those concepts. To demonstrate the predictive value of visual counts and image details, we analyze Kickstarter data. Our results highlight that visual count features are two of the top three predictors of success. Our results also show that simple image detail features such as color matter a lot, and our proposed measures of visual scene elements can also be useful. We supplement our article with R and Python codes that help authors extract image details (https://osf.io/ujnzp/)., Comment: 32 pages, 5 figures

Long horizon sequential manipulation tasks are effectively addressed hierarchically: at a high level of abstraction the planner searches over abstract action sequences, and when a plan is found, lower level motion plans are generated. Such a strategy hinges on the ability to reliably predict that a feasible low level plan will be found which satisfies the abstract plan. However, computing Abstract Plan Feasibility (APF) is difficult because the outcome of a plan depends on real-world phenomena that are difficult to model, such as noise in estimation and execution. In this work, we present an active learning approach to efficiently acquire an APF predictor through task-independent, curious exploration on a robot. The robot identifies plans whose outcomes would be informative about APF, executes those plans, and learns from their successes or failures. Critically, we leverage an infeasible subsequence property to prune candidate plans in the active learning strategy, allowing our system to learn from less data. We evaluate our strategy in simulation and on a real Franka Emika Panda robot with integrated perception, experimentation, planning, and execution. In a stacking domain where objects have non-uniform mass distributions, we show that our system permits real robot learning of an APF model in four hundred self-supervised interactions, and that our learned model can be used effectively in multiple downstream tasks., Comment: To appear in Robotics: Science and Systems 2021

Exploration in novel settings can be challenging without prior experience in similar domains. However, humans are able to build on prior experience quickly and efficiently. Children exhibit this behavior when playing with toys. For example, given a toy with a yellow and blue door, a child will explore with no clear objective, but once they have discovered how to open the yellow door, they will most likely be able to open the blue door much faster. Adults also exhibit this behavior when entering new spaces such as kitchens. We develop a method, Contextual Prior Prediction, which provides a means of transferring knowledge between interactions in similar domains through vision. We develop agents that exhibit exploratory behavior with increasing efficiency, by learning visual features that are shared across environments, and how they correlate to actions. Our problem is formulated as a Contextual Multi-Armed Bandit where the contexts are images, and the robot has access to a parameterized action space. Given a novel object, the objective is to maximize reward with few interactions. A domain which strongly exhibits correlations between visual features and motion is kinemetically constrained mechanisms. We evaluate our method on simulated prismatic and revolute joints., Comment: IEEE International Conference on Robotics and Automation (ICRA) 2020

Background: One of the common causes of sudden cardiac death (SCD) in young people is hypertrophic cardiomyopathy (HCM) and the primary prevention of SCD is with an implantable cardioverter defibrillators (ICD). Concerning the incidence of appropriate ICD therapy and the complications associated with ICD implantation and discharge, patients with implanted ICDs are closely monitored and interrogation reports are generated from clinical consultations. Methods: In this study, we compared the performance of structured device data and unstructured interrogation reports for extracting information of ICD therapy and heart rhythm. We sampled 687 reports with a gold standard generated through manual chart review. A rule-based natural language processing (NLP) system was developed using 480 reports and the information in the corresponding device data was aggregated for the task. We compared the performance of the NLP system with information aggregated from structured device data using the remaining 207 reports. Results: The rule-based NLP system achieved F-measure of 0.92 and 0.98 for ICD therapy and heart rhythm while the performance of aggregating device data was significantly lower with F-measure of 0.78 and 0.45 respectively. Limitations of using only structured device data include no differentiation of real events from management events, data availability, and disparate perspectives of vendor and data granularity while using interrogation reports needs to overcome non-representative keyword/pattern and contextual errors. Conclusions: Extracting phenotyping information from data generated in real-world requires the incorporation of medical knowledge. It is essential to analyze, compare, and harmonize multiple data sources for real-world evidence generation.

Current conversational systems can follow simple commands and answer basic questions, but they have difficulty maintaining coherent and open-ended conversations about specific topics. Competitions like the Conversational Intelligence (ConvAI) challenge are being organized to push the research development towards that goal. This article presents in detail the RLLChatbot that participated in the 2017 ConvAI challenge. The goal of this research is to better understand how current deep learning and reinforcement learning tools can be used to build a robust yet flexible open domain conversational agent. We provide a thorough description of how a dialog system can be built and trained from mostly public-domain datasets using an ensemble model. The first contribution of this work is a detailed description and analysis of different text generation models in addition to novel message ranking and selection methods. Moreover, a new open-source conversational dataset is presented. Training on this data significantly improves the Recall@k score of the ranking and selection mechanisms compared to our baseline model responsible for selecting the message returned at each interaction., Comment: 46 pages including references and appendix, 14 figures, 12 tables; Under review for the Dialogue & Discourse journal

Automatically evaluating the quality of dialogue responses for unstructured domains is a challenging problem. Unfortunately, existing automatic evaluation metrics are biased and correlate very poorly with human judgements of response quality. Yet having an accurate automatic evaluation procedure is crucial for dialogue research, as it allows rapid prototyping and testing of new models with fewer expensive human evaluations. In response to this challenge, we formulate automatic dialogue evaluation as a learning problem. We present an evaluation model (ADEM) that learns to predict human-like scores to input responses, using a new dataset of human response scores. We show that the ADEM model's predictions correlate significantly, and at a level much higher than word-overlap metrics such as BLEU, with human judgements at both the utterance and system-level. We also show that ADEM can generalize to evaluating dialogue models unseen during training, an important step for automatic dialogue evaluation., Comment: ACL 2017

An open challenge in constructing dialogue systems is developing methods for automatically learning dialogue strategies from large amounts of unlabelled data. Recent work has proposed Next-Utterance-Classification (NUC) as a surrogate task for building dialogue systems from text data. In this paper we investigate the performance of humans on this task to validate the relevance of NUC as a method of evaluation. Our results show three main findings: (1) humans are able to correctly classify responses at a rate much better than chance, thus confirming that the task is feasible, (2) human performance levels vary across task domains (we consider 3 datasets) and expertise levels (novice vs experts), thus showing that a range of performance is possible on this type of task, (3) automated dialogue systems built using state-of-the-art machine learning methods have similar performance to the human novices, but worse than the experts, thus confirming the utility of this class of tasks for driving further research in automated dialogue systems., Comment: Accepted to SIGDIAL 2016 (short paper). 5 pages

An anisotropic elasticity tensor can be approximated by the closest tensor belonging to a higher symmetry class. The closeness of tensors depends on the choice of a criterion. We compare the closest isotropic tensors obtained using four approaches: the Frobenius 36-component norm, the Frobenius 21-component norm, the operator norm and the L2 slowness-curve fit. We find that the isotropic tensors are similar to each other within the range of expected measurement errors.

We investigate evaluation metrics for dialogue response generation systems where supervised labels, such as task completion, are not available. Recent works in response generation have adopted metrics from machine translation to compare a model's generated response to a single target response. We show that these metrics correlate very weakly with human judgements in the non-technical Twitter domain, and not at all in the technical Ubuntu domain. We provide quantitative and qualitative results highlighting specific weaknesses in existing metrics, and provide recommendations for future development of better automatic evaluation metrics for dialogue systems., Comment: First 4 authors had equal contribution. 13 pages, 5 tables, 6 figures. EMNLP 2016