Abstract-Humans and animals are capable of learning a new behavior by observing others perform the skill just once. We consider the problem of allowing a robot to do the same -learning from a video of a human, even when there is domain shift in the perspective, environment, and embodiment between the robot and the observed human. Prior approaches to this problem have hand-specified how human and robot actions correspond and often relied on explicit human pose detection systems. In this work, we present an approach for one-shot learning from a video of a human by using human and robot demonstration data from a variety of previous tasks to build up prior knowledge through meta-learning. Then, combining this prior knowledge and only a single video demonstration from a human, the robot can perform the task that the human demonstrated. We show experiments on both a PR2 arm and a Sawyer arm, demonstrating that after meta-learning, the robot can learn to place, push, and pick-andplace new objects using just one video of a human performing the manipulation.
This work presents a novel, comprehensive framework that leverages emerging augmented reality headset technology to enable smart nuclear industrial infrastructure that a human can easily interact with to improve their performance in terms of safety, security, and productivity. Nuclear industrial operations require some of the most complicated infrastructure that must be managed today. Nuclear infrastructure and their associated industrial operations typically features stringent requirements associated with seismic, personnel management (e.g., access control, equipment access), safety (e.g., radiation, criticality, mechanical, electrical, spark, and chemical hazards), security (cyber/physical), and sometimes international treaties for nuclear non-proliferation. Furthermore, a wide variety of manufacturing and maintenance operations take place within these facilities further complicating their management. Nuclear facilities require very thorough and stringent documentation of the operations occurring within these facilities as well as maintaining a tight chain-of-custody for the materials being stored within the facility. The emergence of augmented reality and a variety of Internet of Things (IoT) devices offers a possible solution to help mitigate these challenges. This work provides a demonstration of a prototype smart nuclear infrastructure system that leverages augmented reality to illustrate the advantages of this system. It will also present example augmented reality tools that can be leveraged to create the next generation of smart nuclear infrastructure. The discussion will layout future directions of research for this class of work.
Recent developments in the ability to automatically and efficiently extract natural frequencies, damping ratios, and full-field mode shapes from video of vibrating structures has great potential for reducing the resources and time required for performing experimental and operational modal analysis at very high spatial resolution. Furthermore, these techniques have the added advantage that they can be implemented remotely and in a non-contact fashion. Emerging full-field imaging techniques therefore have potential to allow the identification of the modal properties of structures in regimes that used to be challenging. For instance, these techniques suggest that the high spatial resolution structural identification could be performed on an aircraft during flight using a ground or aircraft-based imager. They also have the potential to identify the dynamics of microscopic systems. In order to realize this capability it will be necessary to develop techniques that can extract full-field structural dynamics in the presence of non-ideal operating conditions. In this work, we develop a framework for the deployment of emerging algorithms that allow the automatic extraction of high-resolution, full-field modal parameters in the presence of non-ideal operating conditions. One of the most notable non-ideal operating conditions is the rigid body motion of both the structure being measured as well as the imager performing the measurement. We demonstrate an instantiation of the framework by showing how it can be used to address, in-plane, translational, rigid body motion. The development of a frame-to-frame keypoint–based technique for identifying full-field structural dynamics in the presence of either rigid body motion is presented and demonstrated in the context of the framework for the deployment of full-field structural identification techniques in the presence of non-ideal operating conditions. It is expected that this framework will ultimately help enable the collection of full-field structural dynamics using measurement platforms including unmanned aerial vehicles, robotic telescopes, satellites, imagers mounted in high-vibration environments (seismic, industrial, harsh weather), characterization of microscopic structures, and human-carried imagers. If imager-based structural identification techniques mature to the point that they can be used in non-ideal field conditions, it could open up the possibility that the structural health monitoring community will be able to think beyond monitoring individual structures, to full-field structural integrity monitoring at the city scale.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.