在SIGGRAPH 2024大会上,英伟达(NVIDIA)公司展示了其在物理AI领域的最新进展,引领了机器人技术发展的下一波浪潮。英伟达创始人兼首席执行官黄仁勋在会上宣布了一系列针对机器人制造商、AI模型开发者和软件制造商的新技术和工具。

英伟达推出了全新的NVIDIA NIM微服务和框架,以及NVIDIA OSMO编排服务,旨在帮助开发、训练和构建下一代人形机器人。这些产品包括用于机器人仿真和学习的NIM微服务,以及用于运行多阶段机器人工作负载的OSMO编排服务。通过这些工具,开发者能够使用少量的人类演示数据来训练机器人,大大节省了时间和成本。

NIM微服务提供了由英伟达推理软件支持的预构建容器,使得部署时间从数周缩短到几分钟。机器人专家可以在NVIDIA Isaac Sim中增强生成式物理AI仿真工作流,并在OpenUSD中生成机器人任务和仿真就绪环境。

NVIDIA OSMO是一项云原生托管服务,它简化了机器人训练和仿真的工作流,将部署和开发周期从数月缩短到一周内。OSMO允许用户对各种任务进行可视化管理,包括合成数据生成、模型训练、强化学习和大规模的机器人测试。

英伟达还展示了其在仿真技术方面的综合生成训练数据的优势,并发布了适用于AI模型的NVIDIA NIM微服务。这些微服务能够生成回答用户查询的OpenUSD语言,这将扩大OpenUSD在机器人、工业设计和工程领域的应用,提高开发者在AI新一轮发展中的能力。

英伟达的这些新工具和技术为机器人开发者提供了先进的数据捕获工作流,使得训练人形机器人的基础模型变得更加高效和经济。通过这些创新,英伟达正在推动整个NVIDIA机器人堆栈的发展,为全球人形机器人开发者和服务提供商开放访问,让他们能够使用最符合其需求的平台、加速库和AI模型。

英语如下:

News Title: NVIDIA SIGGRAPH 2024 Keeps AI Robotics Wave Ahead

Keywords: NVIDIA, Robotics, AI Leadership

News Content: At the SIGGRAPH 2024 conference, NVIDIA Corporation showcased its latest advancements in the field of physical AI, leading the next wave of robot technology development. NVIDIA founder and CEO Jensen Huang announced a series of new technologies and tools aimed at robot manufacturers, AI model developers, and software providers during the event.

NVIDIA introduced the new NVIDIA NIM microservices and framework, along with the NVIDIA OSMO orchestration service, designed to assist in the development, training, and construction of the next generation of humanoid robots. These offerings include NIM microservices for robot simulation and learning, as well as the OSMO orchestration service for running multi-stage robot workloads. Through these tools, developers can train robots using minimal human demonstration data, significantly saving time and costs.

The NIM microservices provide pre-built containers supported by NVIDIA inference software, shortening deployment times from weeks to minutes. Robotics experts can enhance generative physical AI simulation workflows in NVIDIA Isaac Sim and generate robot tasks and simulation-ready environments in OpenUSD.

NVIDIA OSMO is a cloud-native managed service that simplifies the robot training and simulation workflow, reducing the deployment and development cycle from months to within a week. OSMO enables users to manage various tasks through visualization, including synthetic data generation, model training, reinforcement learning, and large-scale robot testing.

NVIDIA also demonstrated its comprehensive advantage in simulation technology for generating training data and released NVIDIA NIM microservices tailored for AI models. These microservices can generate responses in OpenUSD language to user queries, expanding OpenUSD’s applications in robotics, industrial design, and engineering, enhancing developers’ capabilities in the next wave of AI development.

NVIDIA’s new tools and technologies provide robotics developers with advanced data capture workflows, making it more efficient and economical to train the foundational models of humanoid robots. Through these innovations, NVIDIA is driving the development of the entire NVIDIA robotics stack, opening access to global humanoid robot developers and service providers, allowing them to use platforms, acceleration libraries, and AI models that best meet their needs.

【来源】https://www.jiqizhixin.com/articles/2024-07-30-10

Views: 2

发表回复

您的邮箱地址不会被公开。 必填项已用 * 标注