Energy efficiency of flapping flight: understanding and exploitation

扑翼飞行代表了低雷诺数飞行在自然和人造研究中最有效的形式。例如，它允许蝴蝶每年在北美和墨西哥西南部之间迁徙长达3000英里，主要是通过拍打翅膀来利用不稳定的空气动力学机制。其他较大的扑翼物种，如鸟类和蝙蝠，在较高的雷诺数范围内工作，在捕猎、迁徙和躲避捕食者的过程中使用类似的机制来实现飞行。然而，由于非线性运动学和形态的复杂性，以及在处理和充分测量活标本方面的困难，这仍然是最不被理解和利用的自然运动形式之一。该项目提出了一种新颖的风洞平台，用于研究扑翼飞行运动学，使用6轴工业铰接式机械臂，用于跨不同参数空间的扑翼飞行的可重复风洞测试，因此物种。风洞设施是伦敦帝国理工学院的T1风洞，这是一个最大风速高达42米/S的超低湍流风洞设施。这个机翼-机器人-隧道系统的高度机动性使人们能够更深入地了解机翼尖端“扫出”的特定扑翼飞行路线如何允许对非定常结构进行最佳的空气动力学控制，以增加超过典型固定翼失速迎角的升力。将使用同步三分量速度场和表面压力测量来描述非定常气动结构的形成和脱落，目的是将全球流场行为与机翼表面所感受到的压力相关联。这代表了一个“部分可观察的系统”，与各种感兴趣物种的自然感知能力一致--例如，蝙蝠可以通过翼膜上的微小毛发感知气流，大型海鸟的鼻孔被用来在动荡的条件下寻找阵风。因此，该实验寻求了一种生物灵感路径，用于在最小可观测的流场中对运动的机翼进行高效的气动控制。对这些机制的进一步研究提出了一种新的控制方法，通过实时流动观测来驱动手臂-机翼系统，并使用强化学习神经网络来控制机翼的运动学。观测将通过机翼表面压力测量和机翼下的固定速度传感器(或适当的其他仿生传感器系统)实现，控制输入用于改善基于升力最大化框架的扑动行为，以最大化能量效率，例如，增加射程。根据从风洞气流中提取的功率与机器人提供的执行程序所需的能量之间的平衡，寻找一条“最佳扑翼飞行”路径。因此，该项目的总体目标是开发一种可编程的方法来研究扑翼和非定常飞行控制，使人们能够更好地了解自然界中的这种运动，从而为正在进行的节能空气动力学研究提供信息。这种方法还为更高自由度提供了良好的可扩展性，并为使用人工智能进行实时流量控制提供了一种真正新颖的方法。有了这些工具，我们寻求了一种多学科的工程方法来了解扑翼飞行背后的能源效率，这是一条通往英国生物学和动物学学科领域老牌组织的原创途径。该项目允许未来应用于空气动力学、机器人和FWMAV，从而实现向零污染的过渡。

Flapping flight represents the most efficient form of low Reynolds number flight across natural and manmade studies. For example, it allows butterflies to migrate up to 3000 miles between North America and southwestern Mexico annually, largely exploiting unsteady aerodynamic mechanisms via flapping of their wings. Other larger flapping species such as birds and bats, operating in higher Reynolds number regimes, use similar mechanisms to achieve flight during hunting, migration and evasion of predators. However, this remains one of the least understood and exploited natural forms of locomotion, due to the combined complexity of non-linear kinematics and morphology, as well as the difficulty in working with and sufficiently measuring live specimens. This project proposes a novel wind tunnel platform for investigation into flapping flight kinematics, using a 6-axis industrial articulated robotic arm, for repeatable wind tunnel testing of flapping flight across various parameter spaces, and hence species. The wind tunnel facility is the T1 Wind Tunnel at Imperial College London, an ultra-low turbulence facility capable of maximum wind speeds of up to 42 m/s. The high manoeuvrability of this wing-robot-tunnel system enables a deeper understanding of how the specific flapping flight path 'swept out' by the wing tip allows for optimal aerodynamic control of unsteady structures for augmentation of lift beyond typical fixed-wing stall angles of attack. Synchronous 3 component velocity fields and surface pressure measurements will be used to characterise the formation and shedding of unsteady aerodynamic structures, with the aim of correlating global flow field behaviour to the pressure sensed on the surface of the wing. This represents a 'partially observable system', in line with the natural sensing abilities of various species of interest - for example, bats can sense airflow via microscopic hairs on their wing membranes, and the nostrils of large seabirds are used to find gusts to ride during turbulent conditions. The experiment therefore seeks a bioinspired pathway for energy efficient aerodynamic control of a moving wing in a minimally observable flow field.Further investigations into exploitation of these mechanisms propose a novel control approach for actuating the arm-wing system via real-time flow observation, with the wing kinematics controlled using a Reinforcement Learning Neural Network. The observations are to be achieved via surface pressure measurements on the wing and fixed velocity sensors off-wing (or other biomimicking sensor systems as appropriate), with control inputs used to improve the flapping behaviour based on a lift-maximising framework to maximise energy efficiency, for example, to increase the range. An 'optimal flapping flight' path is sought, as determined by power extracted from the wind tunnel flow balanced against energy supplied to perform the procedure from the robot. The overall aim of the project is therefore to develop a programmable approach to investigation of flapping and unsteady flight control, to enable a better understanding of such locomotion in the natural world and hence inform ongoing progress into energy-efficient aerodynamics. This approach also offers good scalability to higher degrees of freedom and a truly novel approach to real-time flow control using artificial intelligence. With these tools, a multi-disciplinary engineering approach to understanding the energy efficiency behind flapping flight is sought, in an original pathway to established UK groups in Biology and Zoological disciplines. The project allows for future application across aerodynamics, robotics and FWMAVs enabling Transition to Zero Pollution.