摘要
由一名新闻记者-机器人与机器学习的工作人员新闻编辑每日新闻-一项关于机器人的新研究-机器人和自动化现在可用。根据NewsRx记者在马萨诸塞州坎布里德的新闻报道，研究表明，“模仿学习(IL)可以从资源密集型模型预测控制器(MPC)中训练计算效率高的传感器运动策略，但它通常需要许多样本，导致训练时间长或鲁棒性有限。为了解决这些问题，我们将IL与考虑过程和感知不确定性的稳健MPC变体结合起来，我们设计了一个数据增强(DA)策略，能够有效地学习基于视觉的策略。这项研究的财政支持来自MURI。新闻记者从麻省理工学院的研究中获得了一句话：“提出的DA方法，命名为Tube-NERF，利用神经辐射场(NeRFs)来生成新的合成图像，并利用鲁棒MPC(Tube)的特性来选择相关的视图并有效地组合相应的动作，我们定制了我们的方法来完成多旋翼的定位和轨迹跟踪任务。通过学习视觉运动策略，通用电气使用机载摄像机的图像作为人体姿势的唯一来源来控制动作。数值评估显示，与目前的IL M方法相比，演示效率提高了80倍，训练时间减少了50%。
Abstract
By a News Reporter-Staff News Editor at Robotics & Machine Learning Daily News Daily News – A new study on Robotics - Robotics and Automation is now available. According to news reporting originating in Cambrid ge, Massachusetts, by NewsRx journalists, research stated, “Imitation learning ( IL) can train computationally-efficient sensorimotor policies from a resource-in tensive model predictive controller (MPC), but it often requires many samples, l eading to long training times or limited robustness. To address these issues, we combine IL with a variant of robust MPC that accounts for process and sensing u ncertainties, and we design a data augmentation (DA) strategy that enables effic ient learning of vision-based policies.” Financial support for this research came from MURI. The news reporters obtained a quote from the research from the Massachusetts Ins titute of Technology, “The proposed DA method, named Tube-NeRF, leverages Neural Radiance Fields (NeRFs) to generate novel synthetic images, and uses properties of the robust MPC (the tube) to select relevant views and to efficiently comput e the corresponding actions. We tailor our approach to the task of localization and trajectory tracking on a multirotor, by learning a visuomotor policy that ge nerates control actions using images from the onboard camera as only source of h orizontal position. Numerical evaluations show 80-fold increase in demonstration efficiency and a 50% reduction in training time over current IL m ethods.”
NETL
NSTL