The First Takeoff of a Biologically Inspired At-Scale Robotic Insect－论文１翻译

The First Takeoff of a Biologically Inspired At-Scale Robotic Insect

第一次生物启发的大规模机器人昆虫的起飞

－－Robert J. Wood,Member, IEEE

Abstract—Biology is a useful tool when applied to engineering challenges that have been solved in nature. Here, the emulous goal of creating an insect-sized, truly micro air vehicle is addressed by first exploring biological principles. These principles give in-sights on how to generate sufficient thrust to sustain flight for centimeter-scale vehicles. Here, it is shown how novel manufactur-ing paradigms enable the creation of the mechanical and aerome-chanical subsystems of a microrobotic device that is capable of Diptera-like wing trajectories. The results are a unique micro-robot: a 6 0mg robotic insect that can produce sufficient thrust to accelerate vertically. Although still externally powered, this mi-cromechanical device represents significant progress toward the creation of autonomous insect-sized micro air vehicles.

Index Terms—Actuators, aerial robotics, biologically inspired robotics, microrobotics.

摘要－在应用于工程时，生物学是一个有用的工具，自然界已经解决的挑战。在这里，创造一个昆虫大小，真正微观的飞行器的宏伟目标，首先通过探索生物学原理来解决。这些原则提供了关于如何产生足够推力以维持厘米级车辆飞行。在这里，它展示了新型制造模式如何能够创建具有双翅类似机翼轨迹的微型机器人装置的机械和机动系统子系统。结果是一个独特的微型机器人：60毫克机器人昆虫可以产生足够的推力以垂直加速。
这种微机械装置虽然仍然由外部供电，但它代表着创造自主式昆虫大小微型飞行器的重大进展。

关键词－执行器，空中机器人，生物启发机器人技术，微型机器人

I. INTRODUCTION

The study of fight began with mankind’s awe and envy of flying organisms. Regarding nature’s smallest fliers, contemporary studies have given detailed insights into the re-markable maneuverability of some flying insects. Insects en-compass the most agile flying objects on earth, including all things man-made and biological. Until recently, the aerody-namic mechanisms by which insects achieve this performance were not understood in the framework of classical aerodynamic theory. Now, through the work of Ellington et al.[1], [2], Dickinson et al. [3]–[5], and others, the complex aerodynamics of a periodic wing stroke at low Reynolds numbers (less than 1000) is understood well enough to be used as a design tool for engineers that wish to recreate these devices.

介绍

对飞行的研究始于人类对飞行生物的敬畏和羡慕。关于自然界最小的飞行器，当代研究已经对一些飞行昆虫的可标记的可操作性给出了详细的见解。昆虫包围地球上最敏捷的飞行物体，包括人造和生物的所有物体。直到最近，昆虫达到这种性能的空气动力学机制还不能在经典的空气动力学理论框架中理解。现在，通过Ellington等人的工作， [1]，[2]，Dickinson et al　[3] - [5]等人，低雷诺数（小于1000）的周期性机翼行程的复杂空气动力学被理解为足以用作希望重新制造这些设备的工程师的设计工具。

But, this understanding is not sufficient to create effective　robotic insects. Novel manufacturing paradigms must be con-sidered concurrently in order to achieve the level of efficiency　and durability that millimeter-scale flying machines will re-quire. The high speed, highly articulated mechanisms that are　necessary to reproduce insect-like wing motions exist on a scale　that is between microelectromechanical systems (MEMS) [6]　and “macro” devices [7]. Thus, a “meso” scale rapid fabrication method, called smart composite microstructures (SCMs),is used to bridge this gap (for details, see [8]). Previous research has yielded concise design rules for the development of　flexure-based micromechanical structures based upon the SCM　process [9], [10]. These design rules form the basis for the de-vice that is described here: an insect-sized flapping-wing micro　air vehicle (MAV).

但是，这种理解不足以创造有效的机器人昆虫。为了达到毫米级飞行器所要求的效率和耐久性水平，必须同时考虑新型制造模式。在微机电系统（MEMS）[6]和“宏观”设备[7]之间存在一个高速，高度连接的机制来重现昆虫般的机翼运动。因此，使用称为智能复合微结构（SCMs）的“中等”尺度快速制造方法来弥补这一差距（详情见[8]）。基于SCM过程[9]，[10]，以前的重新搜索已经为基于挠曲的微机械结构的开发产生了简洁的设计规则。这些设计规则构成了此处描述的设备的基础：昆虫大小的扑翼微型飞行器（MAV）。

Insect-like vehicles have vast potential applications including　search and rescue, hazardous environment exploration, surveil-lance, reconnaissance, and planetary exploration [11]. The pro-totype robotic insect is shown in Fig. 1 (for more details, see　http://micro.seas.harvard.edu).

机翼类昆虫车辆具有广泛的潜在应用，包括搜救，危险环境探测，监视，侦察和行星探索[11]。 原型机器人昆虫如图1所示（更多细节见http://micro.seas.harvard.edu）。

II. INSECT FLIGHT

Insects of the order Diptera generate aerodynamic forces with a three degree-of-freedom wing trajectory that consists of a large wing stroke (that defines the stroke plane), pronation and supina-tion (collectively called wing rotation) about a longitudinal wing axis, and stroke plane deviation [12]–[15]. This discussion will not consider stroke plane deviation: for some hovering Dipteran insects that have a nearly horizontal stroke plane, it does not appear that stroke plane deviation plays a significant role in lift generation [3].This considerably simplifies the analysis of the wing trajectory and the construction of the transmission mechanism. The desired two degree-of-freedom trajectory profile is shown in Fig. 2(a). Dipteran insects generate wing motions using indirect flight muscles that pull on a deformable section of the exoskeleton called the scutum [16]. The wing is connected to the pleural wing process at the interface of the scutum and the exoskeleton. Contractions of the dorsoventral muscles depress the scutum and create the upstroke. Contractions of the dorso-longitudinal muscles shorten the thorax and return the scutum to its initial position, generating the downstroke [see Fig. 2(b) and (c)]. Wing rotation is accomplished by smaller muscles (basalar and subalar) that directly apply a torque to the sclerites connected to the wing hinge [17]. While there has been some debate over the concise mechanisms involved in Dipteran tho-racic mechanics, there are a few clear characteristics of the wing drive system.

昆虫飞行

双翅目昆虫产生具有三个自由度的机翼轨迹的气动力，该机翼轨迹由大型机翼行程（定义行程平面），关于纵向机翼轴线的内旋和回旋（统称为机翼旋转）和行程 平面偏差[12] - [15]。这个讨论不会考虑行程平面飞机的偏差：对于一些盘旋的Dipteran
具有接近水平的行程平面的昆虫，似乎不存在行程平面偏差在升力产生中起重要作用[3]。这大大简化了对机翼轨迹的分析和传动机构的构造。图2（a）显示了所需的两自由度轨迹曲线。双翅目昆虫使用间接飞行肌肉产生机翼运动，这些肌肉拉动外骨骼的变形部分，称为屏障[16]。机翼连接到胸膜的过程通过盾构和外骨骼的界面。背腹肌的收缩抑制了屏幕，并产生了上冲。背部 - 纵向肌肉的收缩使胸部缩短，并使粪便回到其初始位置，产生下行动作[见图2（b）和（c）]。机翼旋转是通过较小的肌肉（basalar和subalar）完成的，它直接将扭矩施加到连接到机翼铰链[1７]的钢板上。虽然关于二分力量动力学机制中涉及的简明机制存在一些争论，但翼驱动系统的一些特征还是很明显的。

1) Insects utilize a mechanical advantage to amplify the wing　stroke.
2) Diptera operate their wing strokes at the natural frequency of the aeromechanical system [13].
3) Some aspects of Dipteran wing trajectories are mechanically “hard-coded” into their morphologies while others
are tunable.
1）昆虫利用机械优势来放大机翼行程。
2）双翅目以航空机械系统的固有频率进行机翼冲程[13]。
3）双翼飞机机翼轨迹的某些方面在机械上被“硬编码”成其形态，而其他方面则是可调的。

Here, we utilize each of these aspects with the creation of aresonant wing-drive system that is mechanically “prepro-
grammed” with a desired baseline trajectory. A parallel is shown in Fig. 2(b)–(e) between Dipteran thoracic morphology (over-simplified) and the robotic version.
在这里，我们利用这些方面的每一个方面创造了机械式“预先编程”且具有期望的基线轨迹的谐振翼驱动系统。 平行显示在图2（b） - （e）之间Dipteran胸部形态（over-简化）和机器人版本。

III. CREATION OF A ROBOTIC INSECT

Insect wing trajectories are the basis for the Harvard Mi-crorobotic Fly. If a robotic device can reproduce the necessary aspects of Dipteran wing motion with a similar wing-beat frequency and mass comparable to actual flies, it should be capable of producing sufficient thrust to fly. This device has four primary mechanical components: the airframe (actuator (flight muscle), transmission (thorax), and airfoils, is shown in Fig. 1(a). The function of each are simple: 1) the frame provides a solid ground to the actuator and transmission while contributing minimal mass; 2) the actuator should provide motion with maximal power density; 3) the transmission must efficiently impedance-match the actuator to the load; and 4) the airfoils must remain rigid to hold shape under large aerodynamic loads. The subtleties of these components are described later.

创建机器人昆虫

昆虫翼轨迹是哈佛大学微型恐龙飞行的基础。如果一个机器人设备可以再现双翅翼运动的必要方面，并具有与实际苍蝇相似的机翼拍频和质量，它应该能够产生足够的飞行推力。该装置有四个主要机械部件：机身（外骨骼），执行器（飞行肌肉），传输（胸部）和机翼，如图1（a）所示。每个功能都很简单：1）空气框架为执行器和变速箱提供了坚实的基础，同时提供最小的质量; 2）执行器应提供最大功率密度的运动; 3）变速箱必须有效地将执行机构与负载进行阻抗匹配; 和4）翼型必须保持刚性以在大的空气动力载荷下保持形状。 这些组件的细微之处在后面描述。

A. Actuation

There is one primary actuator that drives the thorax in a similar configuration to Dipteran dorsoventral muscles, but with　bidirectional force. The actuator chosen for this application is abimorph piezoelectric clamped-free bending cantilever that is optimized for mechanical power delivery and created using SCM [18]. These actuators are chosen because of favorable scal-ability (compared, for example, to electromagnetic motors) and compatibility with the SCM process. Compared to other actuation technologies, piezoelectric actuators typically have high operating stresses and frequencies. However, piezoceramic materials are dense and brittle and typically achieve relatively small strains (hence the need for mechanical amplification). The ultimate quantity of interest for a hover-capable MAV is the power density (at the frequency of interest). For the sake of a mance metric, biological estimates place the body-mass-specific power density for flying insects between 29 W/kg [19] and 40 W/kg [20] and between 80 W/kg [21] and 83 W/kg [22] for the muscles alone. For comparison, the class of actuators used here have demonstrated power densities of 400 W/kg [23]. This is the first example of where a subsystem of the microrobotic fly exceeds the performance of its biological counterpart.

A. 执行

有一个主驱动器驱动与双侧背腹肌类似的配置的胸部，但是具有双向力。为此应用选择的执行器是一个双压电晶体压电式无夹弯曲悬臂梁，针对机械动力传输进行了优化，并使用SCM [18]。选择这些执行器是因为它具有良好的可扩展性（例如与电磁马达相比）以及与SCM过程的兼容性。与其他致动技术相比，压电致动器通常具有较高的操作应力和频率。然而，压电陶瓷材料致密且易碎，并且通常实现相对较小的应变（因此需要机械放大）。悬停MAV的最终利益数量是功率密度（以感兴趣的频率）。为了达到性能指标，生物学估算将飞行昆虫的体质量比功率密度定为29W / kg [19]至40W / kg [20]，80W / kg [21]至83W / kg [22]仅用于肌肉。为了比较，使用的执行器类这里已经证明功率密度为400 W / kg [23]。这是微型机器人飞行子系统超过其生物副本的性能的第一个例子。

The design of the actuator is based upon a laminate plate theory model that describes the stress distribution across a mul-tilayered composite structure (see [18] for details of this model).Since the individual layers are thin, the model is simplified to a reduced-tensor notation. However, all desired characteristics of the actuator are included (e.g., displacement, peak force, band-width, etc.). The actuators are created with the SCM process: first, individual lamina [PbZrTiO3(PZT)-5H and M60J carbon fiber/cyanate ester resin prepreg] are laser-micromachined into desired planform shapes. These layers [see Fig. 3(a)] are then stacked and aligned and put through a controlled cure cycle that regulates temperature, pressure, and time of cure. The resulting actuator is fixed to the airframe proximally and the input to the
transmission distally. Application of an electric field creates a bending moment in the actuator that deflects the transmission [exaggerated in Fig. 3(b)]. These actuators are 40 mg, 12 mm long, and achieve a deflection of greater than ±400μm with a bandwidth greater than 1 kHz.

致动器的设计基于层叠板理论模型，该理论模型描述了多层复合结构的应力分布（有关此模型的详细信息，请参见[18]）。由于单个图层很薄，模型被简化为一个减少张量的符号。然而，包括致动器的所有期望特性（例如，位移，峰值力，宽度等）。执行器使用SCM过程创建：首先，将单独的薄片[PbZrTiO3（PZT）-5H和M60J碳纤维/氰酸酯树脂预浸料]激光微加工成所需的平面形状。然后将这些层[参见图3（a）]堆叠并对齐，并通过调节温度，压力和固化时间的可控固化循环。最终的致动器被固定到机身的近端，并且输入到变速器的远端。施加电场会在致动器上产生弯曲力矩，从而使变速器偏转[图3（b）中夸大了]。这些执行器是40毫克，12毫米长，并实现大于±400微米的偏差并且带宽大于1 kHz。

B. Transmission

Similar to the insect model shown in Fig. 2(b) and (c), the transmission amplifies the actuator motion from a translational input to a rotational output. This is done while impedance-matching the load to the actuator: the system is driven at its fundamental resonance, and thus, the dynamics during normal operation are dominated by the wing loading and the actuator losses. Therefore, for efficient electromechanical transduction, it is imperative that the wing loading (as seen by the actuator) is matched to the internal losses of the actuator.

传输

类似于图2（b）和（c）中所示的昆虫模型，变速器将致动器运动从平移输入放大到旋转输出。这是在将负载与执行器进行阻抗匹配时完成的：系统以其基本共振进行驱动，因此正常运行期间的动态特性由机翼负载和执行器损耗决定。因此，为了高效的机电转换，必须将机翼负载（如执行机构所见）与执行机构的内部损耗相匹配。

At larger scales, such a device could be assembled with gears and slider mechanisms [24]. However, due to unfavorable surface area scaling, such components would result in significant friction losses as the characteristic size is decreased. Instead, flexures are used in place of revolute joints [25] (see Fig. 4). The desired stroke amplitude is ±60◦; thus,for actuator motion of approximately±400μm, we require a transmission thathas a nominal amplification of approximately 2600 rad/m. This is called the transmission ratio and is directly analogous to a gear ratio for these compliant mechanisms. Note that this as-sumes a perfectly compliant transmission. If the stiffness of the transmission is nonzero, the transmission ratio will need to be increased.

在更大的尺度上，这种装置可以与齿轮和滑块机构组装[24]。然而，由于不利的表面积缩放，随着特征尺寸的减小，这些部件将导致显着的摩擦损失。相反，使用弯曲来代替旋转关节[25]（见图4）。所需的行程幅度为±60°; 因此，对于大约±400μm的执行器运动，我们需要一个标称放大率约为2600 rad / m的变速箱。这被称为传动比，并且与这些顺应机构的传动比直接类似。请注意，这是一个完美兼容的传输。如果变速器的刚度不为零，则传动比将需要增加。

There has been significant research into the creation and char-acterization of flexure-based micromechanical devices[26], and these design rules provide the basis for the creation of the transmission. These flexure mechanisms are also created using the SCM paradigm. To create jointed structures, rigid materials (carbon fiber reinforced composite prepregs) are cut as face sheets and polymers (typically polyimide) are used as the flexure. These are again bonded in a controlled cure cycle. This mechanism is shown in Figs. 2(d) and (e) and 5(a).

对基于挠曲的微机械装置的创建和表征进行了重大研究[26]，并且这些设计规则为创造传输提供了基础。这些灵活机制也是使用SCM范例创建的。创建节理结构，刚性材料（碳纤维增强复合材料预浸料坯）作为面板切割，聚合物（通常为聚酰亚胺）用作挠曲件。这些再次以可控的固化周期结合。这种机制如图 2（d）和（e）以及5（a）。

The actuator and transmission directly control the wing stroke. Wing rotation is developed passively with an additional flexure positioned between the output of the transmission and the wing. This flexure is parallel to the spanwise direction and in-cludes joint-stops to avoid overrotation [see Fig. 5(c)]. Thus, this system has three degrees-of-freedom, only one of which is actuated [27] (as opposed to concurrent research on insect-inspired MAVs that attempt to concisely control each degree-of-freedom independently [10]).

执行器和变速箱直接控制机翼行程。机翼旋转被动地发展，并且在变速器的输出端和机翼之间设置有额外的弯曲部。这种弯曲平行于翼展方向，并且包括关节止动以避免过度旋转[见图5（c）]。因此，这个系统有三个自由度，其中只有一个是自动的[27]（而不是同时研究昆虫启发的MAV，试图简洁地控制每个自由度[10]）。

C. Airfoils

The airfoils are designed to match the shape and size of the wings of Syrphid hoverflies. Insect wings have nontrivial anisotropic compliances [28], [29], but the airfoils used here are designed to remain rigid for all expected loading conditions. The airfoils are morphologically similar to insect wings; however, the “veins” consist of 30μm thick ultrahigh modulus carbon fiber reinforced composite beams and the “membrane” is 1.5μm thick polyester. The veins are arranged so that the wing is extremely rigid over the expected range of flight forces. To enable quasi-static passive wing rotation, the rotational inertia must be low such that the rotational resonant frequency is sufficiently high. Assuming underdamped second-order dynamics, an acceptable criterion for quasi-static rotation is that the first 3dB point for rotation occurs above the flapping frequency. The 15 mm wing shown in Fig. 5(b) weighs 400μg and remains un-deformed during the entire stroke [see Fig. 7(a)]. These wings exhibit a remarkably high stiffness-to-weight ratio, which is a second example of the superiority of a micromechanical device over a biological system (due to significant differences between the material properties of chitin and carbon fiber).

翼型

翼型的设计符合Syrphid hoverflies翅膀的形状和尺寸。昆虫翅膀具有非平凡的各向异性顺应性[28]，[29]，但这里使用的翼型设计为在所有预期的载荷条件下保持刚性。翼型在形态上类似于昆虫翅膀; 然而，“静脉”由30μm厚的超高模量碳纤维增强复合材料梁组成，“膜”是1.5μm厚的聚酯。静脉布置成使得机翼在预期飞行力范围内非常坚硬。为了实现准静态被动机翼旋转，旋转惯性必须很低，以便旋转共振频率足够高。假设欠阻尼的二阶动力学，准静态旋转的可接受标准是第一个3dB旋转点发生在扑动频率以上。图5（b）所示的15毫米机翼重400克，在整个行程中保持不变形[见图7（a）]。这些翅膀表现出非常高的刚度 - 重量比，这是微机械装置优于生物系统的第二个例子（由于几丁质和碳纤维的材料特性之间的显着差异）。

IV. RESULTS

Each of the components of the fly are assembled onto the airframe resulting in the structure shown in Fig. 1(b). To give an overall perspective, the finished robotic fly is compared to various hover-capable species in Fig. 6. The mass of each component in the integrated structure is given in Table I.

结果

飞行的每个组件都组装到机身上，形成图1（b）所示的结构。为了给出一个总体的看法，完成的机器人苍蝇与图6中的各种悬浮物种相比较。表1给出了整体结构中每个部件的质量。

To isolate thrust from roll, pitch, and yaw moments, the robotic fly is fixed to taut guide wires that restrict the fly to purely vertical motion (guides are fabricated into the airframe). These wires are “training wheels” that will be incrementally removed in future research as attitude sensing and control are migrated onboard. The wings are driven open loop at the flap-ping resonance to maximize the stroke amplitude. For the MAV shown in Fig. 1, the resonant frequency is 110 Hz. So long as the rotational frequency is sufficiently higher than the flapping resonant frequency, the rotational component will operate quasi-statically. The natural frequency for wing rotation is calculated to be approximately 250 Hz based upon the flexural stiffness of the wing hinge and the rotational inertia of the wing [estimated from a computer-aided design (CAD) model]. Thus, the aerody-namic and inertial loads act to decrease the angle of attack, and the joint stops assure that the wings do not over-rotate. To visu-alize the wing motion, a high-speed video camera was used in two perspectives (lateral and anterior). Sequential frames from these videos are shown in Fig. 7. From the original high-speed video (≈20 frames/period), the wing kinematics are extracted(using custom Matlab software) and shown in Fig. 8. The tra-jectory is nearly identical to that of hovering Dipteran insects that lends credence to the use of passive rotation. However, the primary performance metric is the lift that is generated. This is assessed in two ways. First, the fly is fixed to a custom force sensor and the wings are driven open loop (maximum actuator drive field of 2 Vμm−1). Lift measurement trials start by:1) collecting the zero level of the sensor; 2) starting the wing from rest and allowing transients to decay (for 0.5 s at 10 kHz sample rate); and 3) collecting the force data (again for 0.5 s at 10 kHz corresponding to approximately 50 wing beats at 100 samples/period). Over ten trials, an average lift of 1.14±0.23 mN is measured, corresponding to a thrust-to-weight ratio of approximately 2. Second, the integrated fly is aligned to the guide wires and allowed to freely move in the vertical direction. The wings are driven open loop and the fly ascends, as is shown in Fig. 9. This marks the first (tethered) liftoff of an insect-scale microrobot and validates the use of biological inspiration. Table II lists the key characteristics of the integratedfly.

为了隔离侧倾，俯仰和偏航力矩，机器人飞行被固定到拉紧的导线上，以将飞行限制在纯粹的垂直运动中（导向装置被制造到机身中）。这些导线是“训练轮”，在将来的研究中，随着姿态传感和控制在船上的移动，这些将会逐渐消除。他的机翼在翻板共振时被驱动为开环以最大化行程振幅。对于图1所示的MAV，谐振频率为110 Hz。只要旋转频率足够高于扑动共振频率，旋转分量就会准静态工作。根据机翼铰链的挠曲刚度和机翼的转动惯量（根据计算机辅助设计（CAD）模型估算），机翼旋转的固有频率计算为大约250 Hz。因此，气动和惯性负载起到减小攻角的作用，并且关节停止确保机翼不会过度旋转。为了可视化机翼运动，高速摄像机被用于两个视角（横向和前方）。来自这些视频的连续帧在图7中示出。从原始的高速视频（≈20帧/周期），翼运动学被提取（使用自定义的Matlab软件），如图8所示。轨迹与悬停的双翅目昆虫的轨迹几乎完全相同，这可以使人们相信使用被动旋转。但是，主要性能指标是生成的提升。这通过两种方式进行评估。首先，将苍蝇固定在一个定制的力传感器上，并且机翼被驱动为开环（最大致动器驱动场为2Vμm-1）。提升测量试验开始于：1）收集传感器的零位; 2）从静止启动机翼并允许瞬态衰减（在10kHz采样率下0.5秒）; 和3）收集力数据（再次在10kHz下0.5秒，对应于100个样本/周期的大约50次机翼节拍）。超过10次试验，测得的平均升力为1.14±0.23mN，对应于约2的推重比。其次，整合的苍蝇是对准的导线并允许沿垂直方向自由移动。如图9所示，机翼被驱动为开环并且飞行物上升。这标志着昆虫尺度微型机器人的第一个（栓系）升空，并验证了生物启发的使用。表二列出了集成蝇的主要特征。

V. DISCUSSION

In summary, the performance of the Harvard Microrobotic Fly proves: 1) generating wing trajectories similar to Dipteran
insects is possible with a robotic device and 2) propulsion generated by this microrobot is significant (compared to the body mass). However, these results do not show free flight, integrated sensing and control, or onboard power and electronics. There-fore, the Harvard Microrobotic Fly only represents a solution to the mechanical and aeromechanical components of an autonomous robotic insect. In order to create a fully autonomous flying microrobot, there are two additional research challenges that must be solved: 1) high-efficiency sensing, power conditioning, and control microelectronics and 2) a high energy density power source.

讨论

总之，哈佛微型机器人飞行器的性能证明：1）机器人装置可产生与双翅目昆虫类似的机翼轨迹，以及2）由此微型机器人产生的推进力显着（与体重相比）。但是，这些结果并不显示自由飞行，集成传感和控制，或车载电力和电子设备。因此，哈佛微型机器人飞行只代表自主机器人昆虫的机械和航空机械组件的解决方案。为了创建一个完全自主的飞行微型机器人，还有两个额外的研究挑战必须解决：1）高效率感应，功率调节和控制微电子技术; 2）高能量密度电源。

In previous work, a suite of biologically inspired sensors appropriate for attitude estimation and control has been devel-oped [30]. This suite consists of sub-10 mg mechanoreceptive and photoreceptive sensors. Future development must evolve these into a monolithic solution for low-level (stabilization) sensors. Power and control electronics must also be created to drive the actuators. Both power and control architectures must be efficient, high bandwidth, and eventually occupy a very　small volume of silicon. To compound these requirements, the piezoelectric actuators used in the Harvard Microrobotic Fly require high fields (approximately 2 Vμm−1). Efficient,lightweight boost conversion [31], and drive electronics [32] for these actuators are ongoing research topics.

在以前的工作中，已经开发了一套适合于姿态估计和控制的生物灵感传感器[30]。该套件由低于10毫克的机械感受和感光传感器组成。未来的发展必须将这些发展成为低层（稳定）传感器的整体解决方案。还必须创建电源和控制电子设备来驱动执行器。功率和控制架构都必须高效，高带宽，并最终占据非常小的硅片体积。为了满足这些要求，哈佛微型机器人飞行器中使用的压电致动器需要高场（约2Vμm-1）。高效，轻便的升压转换[31]和驱动电子[32]这些执行器是正在进行的研究课题。

The final component necessary for a fully autonomous fly is a power source appropriate for both long duration flight and short energetic bursts. Based upon the electrical power requirements for the current fly, estimates of overall efficiency for the final version and the best available battery chemistry, it is estimated that fly-sized robots can operate for 5–10 min. However, this expected flight time is based upon the energy density of large-scale batteries. While there is no reason to expect that the chemistry should change with a decrease in physical dimensions, a decrease in size will result in a larger surface area to volume ratio, and thereby, require more packaging material for a given volume of battery. Furthermore, these numbers are for hovering in still air conditions. Gusts or air currents will require more power to remain stationary or follow a prescribed trajectory. The flight time will be incrementally increased by improvements in battery technology, increased propulsive efficiency, and the addition of energy harvesting devices (e.g. solar, vibrational, thermal, etc.).

完全自主飞行所需的最后部件是适合长时间飞行和短能量爆发的动力源。基于当前飞行的电力需求，最终版本的总体效率估计和最佳可用电池化学成分，估计飞行尺寸机器人可以运行5-10分钟。但是，这个预期的飞行时间取决于大型电池的能量密度。虽然没有理由期望化学应该随着物理尺寸的减小而改变，但尺寸的减小将导致更大的表面积与体积比，并且因此对于给定体积的电池需要更多的包装材料。此外，这些数字是为了在静止的空气中徘徊。阵风或气流需要更多的动力来保持静止或遵循规定的轨迹。通过改进电池技术，提高推进效率和增加能量采集设备（例如太阳能，振动，热能等），飞行时间将逐渐增加。

It was noted that the maximum lift-to-weight ratio is approximately 2. However, this is only for the mechanical and　aeromechanical structures. To enable the same specific thrust when power, electronics, and control are migrated onboard, we will need to further maximize the propulsive efficiency. For example, doubling the weight of the fly to 120 mg will require twice the current thrust to maintain a thrust-to-weight ratio of 2. This is a current focus that has begun with empirically verified wing modeling and optimization. It is useful at this point to investigate the mass distribution for the fully autonomous fly. This distribution is shown in Fig. 10. Note that in this distribution, the battery represents approximately 40% of the body mass. For terrestrial mobile robots, greater range can be achieved with a larger, higher capacity power source (within reasonable limits). However, for an aerial robot, this is not true since increased power supply mass will require larger lift forces to sustain flight.This increase in mechanical power translates into a corresponding increase in electrical power, and thus, increasing the battery mass has diminishing returns. The mechanical components in Fig. 10 do not require significant reduction from the current masses given in Table I. With regard to the electrical components, consider that a thinned silicon die with dimensions of 4mm×4mm×250μm will have a mass just under 10 mg(with wiring). To put this in perspective, this chip would have　twice the area of a 8051 microcontroller and potentially 500 000 transistors. For perspective, a Drosophila malanogaster has approximately 300 000 neurons. The development of scale ultralow energy controllers for autonomous microsystems has started for sensor network applications [33].

有人指出，最大升重比约为2。但是，这只适用于机械和航空机械结构。为了在电力，电子和控制在船上进行移动时实现相同的特定推力，我们需要进一步最大化推进效率。例如，将苍蝇的重量加倍至120毫克将需要两倍的电流推力以保持推重比2。这是目前的重点，已经从验证的机翼建模和优化开始。在这一点上有用于调查完全自主飞行的质量分布。这种分布如图10所示。请注意，在此分布中，电池约占体重的40％。对于地面移动机器人，使用更大，更高容量的电源（在合理范围内）可以实现更大的范围。但是，对于空中机器人来说，这不是事实，因为增加的电源质量将需要较大的升力来维持飞行。机械功率的这种增加转化为相应的电功率增加，并且因此增加电池质量具有递减的回报。图10中的机械部件不需要从表I中给出的当前质量显着减少。关于电气元件，考虑到尺寸为4mm×4mm×250μm的薄硅片裸片的质量将接近10mg（带有接线）。从这个角度来看，这款芯片的面积将是8051微控制器的两倍，而且可能有50万个晶体管。为了透视，果蝇malanogaster有大约300 000个神经元。用于自主微系统的毫米级超低能量控制器的开发已经开始用于传感器网络应用[33]。

In earlier sections, there were two allusions to subsystems that outperform biological counterparts. This is by no means universal; there are countless aspects of flying insects that were heretofore unattainable with mechanical recreations. For example, evolution has done a phenomenal job with the integration of actuation, power, and control. However, due in part to recent advances in fabrication techniques (i.e., SCM), this gap is progressively narrowing.

在前几节中，有两个子系统的典故胜过生物学。这绝不是万能的; 飞行昆虫有许多方面都是迄今为止机械娱乐无法实现的。例如，演变在驱动，动力和控制的整合方面做得非常出色。然而，部分原因是制造技术（即SCM）的最新进展，这种差距正在逐渐缩小。

Biological inspiration is ubiquitous in the mechanical subsystems that make up the Harvard Microrobotic Fly. Because of the relative ease of microfabrication allowed by SCM, these micromechanical devices can now be used in the opposite direction to gain insights into biological systems. For example, there are open questions pertaining to the scaling of biological airfoils and their compliances: do performance gains correspond to anisotropic compliances or are airfoil compliances simply due to biological material limitations? This may be a difficult question to answer with biological observation; but, by fabricating microrobotic airfoils with a large space of physical parameters, we can ascertain the role of compliance in lift generation, propulsive efficiency, stability, etc. This forms a sort of “closed loop biological inspiration” that could be a valuable tool for biomechanics researchers.

生物灵感在组成哈佛微型机器人飞行器的机械子系统中无处不在。由于SCM所允许的微制造相对容易，这些微机械装置现在可以用于相反的方向，以深入了解生物系统。例如，有关生物翼型及其顺应性的缩放的开放性问题：性能增益对应于各向异性顺应性还是仅由于生物材料限制导致的翼型顺应性？这可能是一个难以回答的生物观察问题; 但通过制造具有大的物理参数空间的微型机翼翼型，我们可以确定顺应性在升力产生，推进效率，稳定性等方面的作用。这形成了一种“闭环生物学启发”，可能成为生物力学研究人员的宝贵工具。

REFERENCES