ID 原文 译文
24855 对于无约束 L2⁃L∞范数的子问题 II，将L∞范数做近似光滑化处理，并通过梯度下降法求解。 Regarding subproblem-II, which has L2-L∞terms without constraints, we smoothen the L∞term approximately, and employ the gradient descent method to solve it.
24856 交替求解两个子问题至收敛，以求解发射信号。 The transmitted signal is obtained when the solutions of two subproblems are convergent.
24857 仿真实验表明，本文模型在副瓣区域能量抑制上较基于L1范数的区域聚焦照射模型具有更优性能，且本文算法实用性更强。 Numerical experiments reveal that the proposed model has better performance on alleviating the energy in sidelobe regions, and the adopted algorithm is more practical than the method based on L1-norm.
24858 水下机器人的自主快速精确定位是完成海洋资源勘探、目标探测定位与追踪等水下作业任务的前提。 Rapid localization of autonomous underwater vehicles (AUVs) plays an important role in target pursuit tasks.
24859 论文研究基于相对测量的水下机器人主动定位方法，解决存在大的初始定位偏差情况下多水下机器人的快速定位问题。 We study active localization method for AUVs using noisy relative measurement, which achieves the precise position estimate of AUVs as quickly as possible under inaccurate initial estimates.
24860 论文提出包括测量、估计和控制三个模块的多水下机器人快速主动定位框架，降低相对测量误差、初始偏差带来的定位不确定性，同时使多机器人定位具有良好的可扩展性 。 A framework for active localization of AUVs with excellent scalability is proposed, which is composed of measurement module, control module and estimation module.
24861 提出的主动接近信标策略优势在于：被定位机器人与信标的相对几何位置收敛过程中，机器人的定位估计快速指数收敛。 In control module we design the motion strategy for AUV, which makes simultaneous convergence of position estimate and the relative geometric location between AUV and beacon.
24862 利用受噪声干扰的相对测距信息，论文采用强化学习方法实现提出的主动接近信标机动策略。 Using noisy relative measurement, a method based on reinforcement learning is adopted to achieve the motion strategy.
24863 开展的数值仿真实验结果表明：相对于基于圆形轨迹、梳状形轨迹机动策略的定位方法，论文所提方法使得水下机器人定位过程具有更好的快速性和鲁棒性。 The numerical simulation results show that the proposed framework and motion strategy has better rapidity and robustness than the traditional localization method.
24864 目前基于视觉的动态头势识别算法泛化能力弱、识别率低,头戴式传感器的方法经济性、便携性差。 Present vision based on dynamic head gesture recognition algorithms usually have disadvantages in generalization and recognition rate, and head-mounted sensors are expensive and inconvenient.