融合深度学习的无人驾驶多目标检测算法

目标检测是目前计算机视觉领域非常热门的研究方向,是无人驾驶技术重要环节。为提高行车过程中目标检测精度并改善基于单发多目标检测器在训练时出现的正负样本失衡问题,基于车载视频,运用深度学习算法中具有强大性能的SSD模型,通过引入Focal Loss函数设计新的损失函数,解决样本失衡问题;同时在不降低检测速率的情况下,提高检测精度。基于自行采集的车载视频数据集进行实验,结果表明,改进后SSD模型的mAP相较于原始SSD模型提高了3%,达到74%。     

悬停时考虑地面效应的直升机的配平

在单旋翼涵道尾桨直升机的非线性全量飞行动力学数值模型的基础上,提出了求解配平问题的一种混合算法,研究了地面效应对配平的影响。以某直升机为算例,在无地面效应和考虑不同悬停高度时,分别进行了初始参数进行了配平,配平计算结果与地面效应的理论相符合。     

Getting sloshed in outer space

We document some basic facts about how liquids behave in the low gravity of outer space and how this poses challenges to space scientists and engineers who need to design and test systems that handle fluids in outer space while working in the 1-g environment on Earth. N Ananthkrishnan is Director, Coral Digital Technologies (P) Ltd, a Bangalore-based company working in the area of multidisciplinary simulation and system integration of aerospace, naval and surface vehicles, and in the design and development of autonomous unmanned/micro air vehicles.     

Robust Adaptive Control Design for Rotorcraft Unmanned Aerial Vehicles Based on Sliding Mode Approach

This paper presents a nonlinear robust control design method for a generic rotorcraft unmanned aerial vehicle（RUAV）. The control objective is to let the RUAV track some pre-defined time-varying position and headingtrajectories. The proposed controller employs feedback linearization process to realize the dynamic decoupling controland applies adaptive sliding mode control to compensate for the parametric uncertainties and external disturbances.The global asymptotical stability is proved via stability analysis. Compared with the cascaded controller, the proposedcontroller demonstrates a superior tracking performance and robustness through numerical simulation in the presenceof parametric uncertainties and unknown disturbances.     

Life extending control of helicopter based on dynamic inversion

The purpose of using life extending control for Black Hawk UH-60 helicopter is to make a trade-off between the handling qualities and the service life of critical components. An increase in service life span results in enhanced safety and the reduction in maintenance costs. This paper presents a design methodology of life extending control for structural durability and high performance of mechanical system, which is based on an explicit dynamic inversion control scheme. A real-time nonlinear fatigue crack growth model is built to predict fatigue damage resulting from the impact of cyclic bending stress on rotor shaft, which serves as an indicator of service life. The 4-axis gain- scheduled flight controller, whose gains are adjusted as a function of damage and flight velocity, is designed to regu- late roll attitude, pitch attitude, vertical velocity and yaw rate. The nonlinear system simulation results show that the responses can meet the requirements on ADS-33 Level 1 handling qualities and that the 4-axis decoupling control is realized. As the damage increases, the tracking performance is slightly degraded, which results in smaller transients in bending moment response.     

智能航模遥控直升机电路的设计与智能控制

设计了智能航模遥控直升机的硬件电路,并对电路的设计要点做出分析;给出了程序控制框架,通过读取航模直升机上姿态传感器的数据,经过卡尔曼滤波后计算出直升机在空中的综合姿态,结合PID算法,实现航模遥控直升飞机的智能控制.该设计在智能航模直升飞机中得以验证,实验表明该系统降低了航模直升飞机的操作难度,减少了摔机的概率.     

科学探究能力的多元化评价——以美国康涅狄格州的科学探究能力评价为例

在美国康涅狄格州公立高中的统测中,科学探究能力评价作为单项进行,它以学生喜闻乐见的当地资源为探究题材,采用独立与合作探究相结合、自评与他评相结合的多元化方式,有利于激发学生的探究兴趣和潜能,展现学生的真实探究能力.     

某型直升机总桨距操纵杆振动故障研究

振动是直升机飞行中常发生的故障之一,严重影响直升机的飞行安全。某型直升机在试飞过程中,连续出现多架左座总桨距操纵杆振动偏大的故障,给正在开展的试飞工作带来了极大的困难。针对该故障通过测振分析、试验和对比其它同型号直升机,准确判断出导致问题发生的原因,从而采取相应的改进措施并进行试飞验证,最终使得直升机总桨距操纵杆振动水平大大降低,消除了质量安全隐患,确保了该型直升机的操纵可靠性,并为后续类似问题提供了解决思路。     

新千年美国高校“直升机父母”过度参与行为探析

新千年伊始,经济发展迟缓、失业增加、学费高涨等因素促使美国学生的家长倍加关注子女的教育,并且其关注范围从中小学延伸到了大学。许多家长过度参与子女的大学生活,甚至直接干涉大学的教育管理事务,影响了学校正常的运行秩序,也对子女自我管理能力的养成产生了不利影响。针对这种现象,校方也采取各种方式应对,并收到了一定的效果。     

一种基于电话网的远程监测系统及其关键技术

介绍了基于电话网的远程监测系统的基本组成及其系统结构,提出了监控系统几个关键技术的实现方法:利用串口通信开发技术,实现了对机房环境温湿度等的实时测量和自动监测。