本文提出了一种使用对象检测网络在汽车雷达数据上学习对象的笛卡尔速度的方法。提出的方法是在为速度生成自己的训练信号方面进行的。标签仅用于单帧，定向边界框（OBB）。不需要昂贵的笛卡尔速度或连续序列的标签。一般的想法是在不使用单帧OBB标签的情况下预先培训对象检测网络，然后利用网络的OBB预测未标记的数据进行速度训练。详细说明，使用预测的速度以及未标记框架的更新OBB之间的距离和标记框架的OBB预测之间的距离，将网络对未标记帧的OBB预测更新为标记帧的时间戳，用于生成一个自我的预测。监督速度的训练信号。检测网络体系结构由一个模块扩展，以说明多次扫描的时间关系和一个模块，以明确表示雷达的径向速度测量值。仅首次训练的两步方法使用OBB检测，然后使用训练OBB检测和速度。此外，由雷达径向速度测量产生的伪标记的预训练引导Bootstraps本文的自我监督方法。公开可用的Nuscenes数据集进行的实验表明，所提出的方法几乎达到了完全监督培训的速度估计性能，但不需要昂贵的速度标签。此外，我们优于基线方法，该方法仅使用径向速度测量作为标签。

本文介绍了新型混合体系结构，它们结合了基于网格的处理，以改善基于雷达对象检测网络的检测性能和方向估计。纯粹基于网格的检测模型在输入点云的鸟眼视图（BEV）投影上运行。这些方法通过离散的网格分辨率损失了详细信息的损失。这特别适用于雷达对象检测，其中相对粗糙的网格分辨率通常用于解释雷达点云的稀疏性。相反，基于点的模型不会受到此问题的影响，因为它们在没有离散化的情况下处理点云。但是，它们通常表现出比基于网格的方法更差的检测性能。我们表明，基于点的模型可以在网格渲染之前提取邻域功能，利用点的确切相对位置。这对于随后的基于网格的卷积检测主链具有重大好处。在公共Nuscenes数据集的实验中，我们的混合体系结构在检测性能方面取得了改进（汽车类的地图比次要的雷达范围提交比仅限雷达提交的地图高19.7％）和方向估计值（11.5％的相对方向改善）比以前文献的网络相比。

Springs are efficient in storing and returning elastic potential energy but are unable to hold the energy they store in the absence of an external load. Lockable springs use clutches to hold elastic potential energy in the absence of an external load but have not yet been widely adopted in applications, partly because clutches introduce design complexity, reduce energy efficiency, and typically do not afford high-fidelity control over the energy stored by the spring. Here, we present the design of a novel lockable compression spring that uses a small capstan clutch to passively lock a mechanical spring. The capstan clutch can lock up to 1000 N force at any arbitrary deflection, unlock the spring in less than 10 ms with a control force less than 1 % of the maximal spring force, and provide an 80 % energy storage and return efficiency (comparable to a highly efficient electric motor operated at constant nominal speed). By retaining the form factor of a regular spring while providing high-fidelity locking capability even under large spring forces, the proposed design could facilitate the development of energy-efficient spring-based actuators and robots.

Springs can provide force at zero net energy cost by recycling negative mechanical work to benefit motor-driven robots or spring-augmented humans. However, humans have limited force and range of motion, and motors have a limited ability to produce force. These limits constrain how much energy a conventional spring can store and, consequently, how much assistance a spring can provide. In this paper, we introduce an approach to accumulating negative work in assistive springs over several motion cycles. We show that, by utilizing a novel floating spring mechanism, the weight of a human or robot can be used to iteratively increase spring compression, irrespective of the potential energy stored by the spring. Decoupling the force required to compress a spring from the energy stored by a spring advances prior works, and could enable spring-driven robots and humans to perform physically demanding tasks without the use of large actuators.

In this paper, we present an adjustable-equilibrium parallel elastic actuator (AE-PEA). The actuator consists of a motor, an equilibrium adjusting mechanism, and a spring arranged into a cylindrical geometry, similar to a motor-gearbox assembly. The novel component of the actuator is the equilibrium adjusting mechanism which (i) does not require external energy to maintain the equilibrium position of the actuator even if the spring is deformed and (ii) enables equilibrium position control with low energy cost by rotating the spring while keeping it undeformed. Adjustable equilibrium parallel elastic actuators resolve the main limitation of parallel elastic actuators (PEAs) by enabling energy-efficient operation at different equilibrium positions, instead of being limited to energy-efficient operation at a single equilibrium position. We foresee the use of AE-PEAs in industrial robots, mobile robots, exoskeletons, and prostheses, where efficient oscillatory motion and gravity compensation at different positions are required.

Text classification of unseen classes is a challenging Natural Language Processing task and is mainly attempted using two different types of approaches. Similarity-based approaches attempt to classify instances based on similarities between text document representations and class description representations. Zero-shot text classification approaches aim to generalize knowledge gained from a training task by assigning appropriate labels of unknown classes to text documents. Although existing studies have already investigated individual approaches to these categories, the experiments in literature do not provide a consistent comparison. This paper addresses this gap by conducting a systematic evaluation of different similarity-based and zero-shot approaches for text classification of unseen classes. Different state-of-the-art approaches are benchmarked on four text classification datasets, including a new dataset from the medical domain. Additionally, novel SimCSE and SBERT-based baselines are proposed, as other baselines used in existing work yield weak classification results and are easily outperformed. Finally, the novel similarity-based Lbl2TransformerVec approach is presented, which outperforms previous state-of-the-art approaches in unsupervised text classification. Our experiments show that similarity-based approaches significantly outperform zero-shot approaches in most cases. Additionally, using SimCSE or SBERT embeddings instead of simpler text representations increases similarity-based classification results even further.

The neural transducer is an end-to-end model for automatic speech recognition (ASR). While the model is well-suited for streaming ASR, the training process remains challenging. During training, the memory requirements may quickly exceed the capacity of state-of-the-art GPUs, limiting batch size and sequence lengths. In this work, we analyze the time and space complexity of a typical transducer training setup. We propose a memory-efficient training method that computes the transducer loss and gradients sample by sample. We present optimizations to increase the efficiency and parallelism of the sample-wise method. In a set of thorough benchmarks, we show that our sample-wise method significantly reduces memory usage, and performs at competitive speed when compared to the default batched computation. As a highlight, we manage to compute the transducer loss and gradients for a batch size of 1024, and audio length of 40 seconds, using only 6 GB of memory.

Nowadays, the PQ flexibility from the distributed energy resources (DERs) in the high voltage (HV) grids plays a more critical and significant role in grid congestion management in TSO grids. This work proposed a multi-stage deep reinforcement learning approach to estimate the PQ flexibility (PQ area) at the TSO-DSO interfaces and identifies the DER PQ setpoints for each operating point in a way, that DERs in the meshed HV grid can be coordinated to offer flexibility for the transmission grid. In the estimation process, we consider the steady-state grid limits and the robustness in the resulting voltage profile against uncertainties and the N-1 security criterion regarding thermal line loading, essential for real-life grid operational planning applications. Using deep reinforcement learning (DRL) for PQ flexibility estimation is the first of its kind. Furthermore, our approach of considering N-1 security criterion for meshed grids and robustness against uncertainty directly in the optimization tasks offers a new perspective besides the common relaxation schema in finding a solution with mathematical optimal power flow (OPF). Finally, significant improvements in the computational efficiency in estimation PQ area are the highlights of the proposed method.

我们说明了一种可以利用用于构建先验遵守身体定律的神经网络的方法。我们从简单的单层神经网络（NN）开始，但避免选择激活功能。在某些条件和无限宽度极限下，我们可以应用中央限制定理，NN输出变为高斯。然后，我们可以通过依靠高斯过程（GP）理论来调查和操纵极限网络。据观察，作用于GP的线性操作员再次产生GP。对于定义微分方程并描述物理定律的差分运算符也是如此。如果我们要求GP或等效地遵守物理定律，那么这将产生与GP的协方差函数或内核的方程式，其解决方案等效地限制了模型以遵守物理定律。然后，中央限制定理建议可以通过选择激活函数来构建NNS来遵守物理定律，从而使它们在无限宽度极限中匹配特定的内核。以这种方式构建的激活函数可以保证NN先验遵守物理学，直到非限制网络宽度的近似误差。讨论了均匀的1D-螺旋方程的简单示例，并将其与天真的内核和激活进行了比较。

尽管在最近的研究中，冷水珊瑚的分布模式（例如paragorgia achorea）受到了越来越多的关注，但对它们的原位活性模式知之甚少。在本文中，我们使用机器学习技术检查了灰木杆菌中的息肉活动，以分析从挪威Stjernsund部署的自主登录机群集获得的高分辨率时间序列数据和照片。本文得出的模型的互动说明是作为补充材料提供的。我们发现，珊瑚息肉扩展程度的最佳预测指标是当前方向，滞后为三个小时。与水流无直接相关的其他变量（例如温度和盐度）提供了更少的有关息肉活动的信息。有趣的是，可以通过对测量位点上方的水柱中的层流进行采样，而不是通过对珊瑚的直接流中的更湍流流进行采样。我们的结果表明，灰木息肉的活性模式受Stjernsund的强潮流状态的控制。看来，木托氏菌对环境当前状态的较短变化没有反应，而是根据潮汐周期本身的大规模模式来调整其行为，以优化营养的吸收。