1,721,040 research outputs found
Recommended from our members
VR teleoperation interface for learning loco-manipulation of humanoid robots
Full text linkOur world is designed by humans, for humans. This makes humanoid robots the most suitable general-purpose platform to automate repetitive or dangerous tasks done by people. However, due to the complex dynamics and high degrees-of-freedom of humanoid robots as well as the shortage of demonstration data, research in robot learning for humanoids is scarce. To address these challenges, I present a VR interface named TRILL (TeleopeRation Interface for Learning Loco-manipulation) to collect human demonstrations for humanoid robots in both simulation and reality. The demonstrations are then used to train a baseline Imitation Learning algorithm that uses an underlying controller to abstract away the complexity of whole-body control. I further propose that by embedding this data collection mechanism in VR video games, we can amass a large-scale dataset of high quality human demonstrations that can drive the development of future autonomous humanoids. To illustrate the feasibility of this idea, we collect a small dataset on toy tasks in simulation and real robot using the VR interface. We then show that the trained policy can be deployed in simulation with a reasonable success rate. A video demo of the VR teleoperation can be found here: https://youtu.be/PNZTwtcRhVU.Computer Scienc
Recommended from our members
Building digital twins of articulated objects and scenes through interactive perception
Full text linkPerceiving and interacting with complex household environments with human-level robustness and flexibility is an open challenge in embodied AI and robotics. Despite significant progress in AI-powered robotics, manipulation algorithms currently lack the necessary generalization and robustness required for widespread deployment. We aim to close this gap by building digital twins of the real world, which can be instantiated in physical simulation for developing and validating mobile manipulation algorithms. Our work proposed to build the digital twins of objects and indoor scenes from the robot's embodied experience in the real world. Experiments show that our approach can be applied to the real scenario and reconstruct an accurate articulation model without any modification.Computer Scienc
Recommended from our members
Towards convergent offline reinforcement learning
Full text linkReinforcement learning (RL) is a leading method for automated sequential decision-making. However, RL is rarely used in the real world. This neglect is due to instability and inefficiency, often even occurring on benchmark problems. Conversely, the sweeping real-world impact of supervised learning (SL) began years ago. Given how much more powerful RL can be than SL, the value of making RL stable and efficient would be immense. Since we know SL is stable and efficient, one candidate for making RL more stable and efficient is to approach RL more like we approach SL. Indeed, RL algorithms with SL-like convergence guarantees already help stability, and RL algorithms designed to learn from offline datasets (static datasets) already help sample efficiency. In this thesis, we find that combining the two performs far better than expected, and we attempt to figure out why. We conclude with ideas for fixing one of the most straightforward convergent RL algorithms.Computer Scienc
Recommended from our members
A self-supervised approach to data curation for large-scale imitation learning
Full text linkImitation learning advances robot capabilities by enabling the acquisition of diverse behaviors from human demonstrations. However, large-scale datasets used for policy training often introduce substantial variability in quality, which can negatively impact performance. As a result, automatically curating data by filtering datasets to improve quality becomes essential. Existing robotic curation approaches either rely on impractical manual annotations or function at a coarse granularity, such as the dataset or trajectory level, failing to account for the quality of individual state-action pairs. To address this, we propose Scizor, a self-supervised data curation framework that filters out low-quality state-action pairs to improve the performance of imita- tion learning policies. Scizor targets two complementary sources of low-quality data: suboptimal data, which hinder learning with undesirable actions, and redundant data, which dilute training with repetitive patterns. Scizor leverages a self-supervised task progress predictor for suboptimal data to remove samples lacking task progression, and a deduplication module operating on joint state-action representation for sam- ples with redundant patterns. Empirically, we show that Scizor enables imitation learning policies to achieve higher performance with less data, yielding an average improvement of 14.8% across multiple benchmarks.Computer Scienc
Going Beyond Counting First Authors in Author Co-citation Analysis
Full text linkThe present study examines one of the fundamental aspects of author co-citation analysis (ACA) - the way co-citation
counts are defined. Co-citation counting provides the data on which all subsequent statistical analyses and mappings
are based, and we compare ACA results based on two different types of co-citation counting - the traditional type that
only counts the first one among a cited work's authors on the one hand and a non-traditional type that takes into
account the first 5 authors of a cited work on the other hand. Results indicate that the picture produced through this non-traditional author co-citation counting contains more coherent author groups and is therefore considerably clearer. However, this picture represents fewer specialties in the research field being studied than that produced through the traditional first-author co-citation counting when the same number of top-ranked authors is selected and analyzed. Reasons for these effects are discussed
Recommended from our members
LightMARL : smart swarm coordination in urban spaces
Full text linkMulti-Agent Reinforcement Learning (MARL) systems generally require substantial computational resources and high-bandwidth communication, which restricts their deployment to centralized cloud infrastructures. This thesis introduces LightMARL, an optimization framework that facilitates effective multi-agent coordination on resource-constrained edge devices while ensuring coordination quality and real-time performance. LightMARL tackles three primary challenges: computational efficiency, communication overhead, and scalability. To improve computation, the framework utilizes neural network quantization, structured pruning, and knowledge distillation tailored for multi-agent policy optimization (MAPPO). These methods decrease model size and computational demands while maintaining essential coordination behaviors. To enhance communication efficiency, vector quantization combined with delta compression, attention-driven selective information sharing, and predictive protocols minimize bandwidth usage and latency. The framework is structured as a modular Python system incorporating C++ components, supporting deployment on Nvidia Jetson platforms. Experimental validation in simulated environments, including drone swarms, vehicle platooning, and sensor networks, demonstrates the framework’s effectiveness. This study illustrates that complex multi-agent coordination is possible on edge devices, enabling new applications in autonomous functionality.Computer Scienc
Recommended from our members
Deep learning for medical imaging in developing nations
Full text linkDeep learning research and innovation have primarily been focused on high- income countries with abundant imaging data, IT infrastructures, local equipment, and clinical expertise. While the application of Deep Learning (DL) in medical imag- ing has gained popularity, particularly for its ability to perform on par with medical experts and bring new promises to the field of medicine, progress in limited-resource environments where medical imaging is crucial has been relatively slow. For instance, in Sub-Saharan Africa, the rate of perinatal mortality, which refers to baby deaths during pregnancy or the first week due to healthcare/maternal issues, is very high due to limited access to antenatal screening. In these countries, deep learning models could be implemented to help clinicians acquire fetal ultrasound planes for the diagnosis of fetal abnormalities. Although the latest deep learning models have been able to identify standard fetal planes, there is no evidence of their ability to generalize in settings with limited resources, such as areas with restricted access to high-end ultrasound equipment and ultrasound data, or different populations. How can breakthroughs in medical deep learning research be disseminated to the global community? Moreover, how can individuals outside of America benefit from and leverage its value? In order for deep learning models to be adopted in developing countries, there is a need for greater efficiency, and we also require more robust and privacy-preserving models to make them practical. With these questions in mind, my thesis centers on the use of deep learning in healthcare for developing nations. Specifically, I explore and propose efficient, privacy-preserving, and robust machine learning techniques to enhance the efficacy of deep learning models in healthcare. Additionally, I conduct a review of the current state of healthcare in developing regions around the world and consider how deep learning can be utilized to improve patient outcomes and support clinicians.Computer Scienc
Variations on the Author
Full text link“Variations on the Author” discusses two of Eduardo Coutinho’s recent films (Um Dia na Vida, from 2010, and Últimas Conversas, posthumously released in 2015) and their contribution to the general question of documentary authorship. The director’s filmography is characterized by a consistent yet self-effacing form of authorial self-inscription: Coutinho often features as an interviewer that rather than express opinions propels discourses; an interviewer that is good at listening. This mode of self-inscription characterizes him as an author who is not expressive but who is nonetheless markedly present on the screen. In Um Dia na Vida, however, Coutinho is completely absent form the image, while Últimas Conversas, on the contrary, includes a confessional prologue that moves the director from the margins to the center of his films. This article examines the ways in which these works stand out in the filmography of a director who offers new insights into the notion of cinematic authorship
Appropriate Similarity Measures for Author Cocitation Analysis
Full text linkWe provide a number of new insights into the methodological discussion about author cocitation analysis. We first argue that the use of the Pearson correlation for measuring the similarity between authors’ cocitation profiles is not very satisfactory. We then discuss what kind of similarity measures may be used as an alternative to the Pearson correlation. We consider three similarity measures in particular. One is the well-known cosine. The other two similarity measures have not been used before in the bibliometric literature. Finally, we show by means of an example that our findings have a high practical relevance.information science;Pearson correlation;cosine;similarity measure;author cocitation analysis
- …
