Inflatable Humanoid Cybernetic Avatar for Physical Human-Robot Interaction
用于人机交互的可充气仿人控制论虚拟人

In a digital twin, a humanoid robot can be the counterpart of a simulated agent in the real world. In addition, a human, virtual avatar, and avatar robot might constitute digital triplets. We propose an inflatable cybernetic avatar (CA) with a humanoid upper body using an inflatable structure that can represent gestures. This inflatable CA is much lighter, safer, and cheaper than conventional humanoid robots and can be folded when deflated. These properties are ideal for physical human-robot interaction (pHRI) and allow real-time collection of human behavior through interaction. In the experiment, basic movements such as nodding and raising arms were measured using motion capture systems. This paper demonstrates the proposed inflatable CA in a hybrid event. We also conducted an experiment to measure the touch interactions using tactile sensors attached to the fabric of the inflatable part. A psychologically secure inflatable humanoid CA is a promising platform for physical interaction experiments.
在数字孪生中，仿人机器人可以是现实世界中模拟代理的对应物。此外，人类、虚拟化身和化身机器人可能构成数字三胞胎。我们提出了一种可充气的控制论化身（CA），它的上半身是人形的，采用可充气的结构，可以表现手势。这种充气 CA 比传统的仿人机器人更轻、更安全、更便宜，而且在放气后可以折叠。这些特性非常适合物理人机交互（pHRI），并可通过交互实时收集人类行为。在实验中，使用动作捕捉系统测量了点头和举臂等基本动作。本文在一个混合活动中演示了所提出的充气 CA。我们还进行了一项实验，使用附在充气部件织物上的触觉传感器测量触摸互动。心理安全的充气仿人 CA 是一个很有前景的物理交互实验平台。

Keywords: soft robotics, inflatable robot, cybernetic avatar, physical human-robot interaction (pHRI)
关键词：软机器人、充气机器人、控制论化身、物理人机交互（PHRI）

Robots act as a link between the cyber and real worlds. The potential of humanoid robots is not limited to factory automation; they could also be used as remotely controlled robots . The related terms are telepresence and avatar robots. Education is a field in which telepresence robots can be used. Moreover, there are promising directions for enhancing virtual transnational education [3], for example.
机器人是连接网络世界和现实世界的纽带。仿人机器人的潜力不仅限于工厂自动化，它们还可用作远程控制机器人 。与之相关的术语是远程呈现和化身机器人。远程呈现机器人可用于教育领域。此外，加强虚拟跨国教育也是大有可为的方向[3]。

Teleoperated robots that can represent the appearance and movements of an operator are often referred to as avatar robots. Furthermore, telexistence refers to technologies that enable action and sensation through an avatar robot and generate a solid sense of existence for the robot operator [4]. In an inclusive society, it is important to involve avatar robots in social communication, particularly during a pandemic. One of the most advanced avatar robots is the torso, upper limbs, hands, and head, operated using a head-mounted display and haptic feedback based on the concept of telexistence [5]. Compared with simple telepresence robots with tablets and mobile platforms, avatar robots tend to be more complex and closer to humanoid robots. The combination of a motion capture system and a dual-armed robot reproduces human arm movements well [6,7]. Dual-arm robots are used for physical human-robot interactions (pHRI), such as assisting in changing clothes [8]. A study to complement human-robot interaction (HRI) in a robot competition by using a virtual reality (VR) platform is an exciting direction .
能够代表操作者外观和动作的远程操作机器人通常被称为化身机器人。此外，"远程感知 "指的是能够通过化身机器人实现行动和感知，并为机器人操作者带来真实存在感的技术[4]。在一个包容性社会中，让化身机器人参与社会交流非常重要，尤其是在大流行病期间。最先进的化身机器人之一是躯干、上肢、手和头，通过头戴式显示器和基于遥感概念的触觉反馈进行操作[5]。与使用平板电脑和移动平台的简单远程呈现机器人相比，虚拟化身机器人往往更为复杂，更接近仿人机器人。动作捕捉系统与双臂机器人的结合可以很好地再现人类手臂的动作[6,7]。双臂机器人可用于物理人机交互（pHRI），如协助换衣服[8]。利用虚拟现实（VR）平台对机器人竞赛中的人机交互（HRI）进行补充研究是一个令人兴奋的方向 。
Humanoid robots used in HRI are more expensive than VR avatars; furthermore, preparing multiple robots of various shapes and sizes is difficult. Research on humanoid robots that physically hug humans has proposed design guidelines for detecting human gestures and generating autonomous robot behavior [10]. The importance of adding softness to a rigid robot has also been addressed in the context of pHRI [11]. Without robotic avatars that are safe and capable of making many copies, it is unlikely that robots would be fully utilized in cyber-physical space. Therefore, we recognized the need for a new robotic platform for pHRI .
HRI 中使用的仿人机器人比 VR 化身更为昂贵；此外，准备多个不同形状和大小的机器人也很困难。关于与人类身体拥抱的仿人机器人的研究提出了检测人类手势和生成机器人自主行为的设计准则[10]。在 pHRI 的背景下，为刚性机器人增加柔软度的重要性也得到了探讨 [11]。如果没有既安全又能多次复制的机器人化身，机器人就不可能在网络物理空间中得到充分利用。因此，我们认为 pHRI 需要一个新的机器人平台。

Here, we explored a novel avatar robot form that differs from articulated humanoid robots. Related research includes using body structures that differ from the human body. For example, one study involved projecting a facial image onto a drone equipped with a balloon-shaped screen [12]. Considering the advantages of safety, light weight, and freedom of shape, this study focused on inflatable robots [13, 14]. The best known concept of an inflatable robot is Baymax from the film Big Hero 6 by Walt Disney Animation Studios, and an inflatable exterior similar to Baymax has been proposed [15]. The soft appearance of inflatable robots has also been used to develop robotic companions for children with neurodevelopmental disorders [16]. A study of a touch interface that combines a cylindrical inflatable with a mobile platform has also been carried out [17]. Prototype examples of rigid telepresence robots with wire-driven inflatable arms have been developed [18]. Studies on robotizing stuffed animals with strings/tendons are also close in scope to inflat-
在这里，我们探索了一种有别于铰接式仿人机器人的新型化身机器人形态。相关研究包括使用不同于人体的身体结构。例如，一项研究涉及将面部图像投射到装有气球形屏幕的无人机上[12]。考虑到安全、重量轻和形状自由等优点，这项研究重点关注充气机器人[13, 14]。最著名的充气机器人概念是迪斯尼动画工作室出品的电影《大英雄 6》中的贝宝（Baymax），也有人提出了与贝宝类似的充气外观[15]。充气机器人的柔软外观也被用于为神经发育障碍儿童开发机器人伴侣[16]。此外，还对圆柱形充气机器人与移动平台相结合的触摸界面进行了研究 [17]。还开发了带有线驱动充气臂的刚性远程呈现机器人原型[18]。关于用绳索/肌腱使毛绒动物机器人化的研究，在范围上也与充气机器人接近[19]。

Fig. 1. Overview of the inflatable cybernetic avatar. The deformable soft body can vary the possible gestures depending on the number of actuators and attachment points of the wires.
图 1.可充气的控制论虚拟人概述。可变形的软体可根据致动器的数量和导线的连接点来改变可能的手势。
able robots. [19, 20]. 能机器人。[19, 20].
This study proposed an inflatable cybernetic avatar. Here we combined a humanoid inflatable robot with a mobile platform. It is based on a study of the joint mechanisms of inflatable robots [21]. Cybernetic avatars (CAs) include robotic and three-dimensional (3D) graphic avatars and a group of technologies that augment people's physical, cognitive, and perceptual capabilities. Several research groups have been working on CA development and social implementation [22]. The contributions of this paper were to propose the use of an inflatable robot in the context of pHRI and to show the method of designing actuation and sensing inherent to inflatable structures. Moreover, we successfully demonstrated the proposed robot in a hybrid event and preliminary physical interaction experiment.
这项研究提出了一种可充气的控制论化身。在这里，我们将仿人充气机器人与移动平台相结合。它基于对充气机器人关节机制的研究[21]。控制论化身（CA）包括机器人化身和三维（3D）图形化身，以及一组增强人的身体、认知和感知能力的技术。一些研究小组一直致力于 CA 的开发和社会实施 [22]。本文的贡献在于提出了在 pHRI 中使用充气机器人的建议，并展示了充气结构固有的执行和传感设计方法。此外，我们还在一次混合活动和初步物理交互实验中成功演示了所提议的机器人。

The central idea of an inflatable CA is simple: a flexible, empty, and inflatable robot as a physical avatar. Although rigid and dual-armed robots have superior work performance, they are heavy, expensive, and dangerous to contact. In addition, they lack the flexibility of virtual avatars. The proposed inflatable CA is lightweight and inexpensive, thus, easily copied and customized, similar to a digital avatar. In addition, inflatables can be very compact when deflated, even if they are not as compact as digital data. In the proposed prototype, the teleoperation and mobile systems were adapted from an existing telepresence robot (Fig. 1). The human-shaped inflatable upper body was added as a wearable device for a simple telepresence robot to constitute the humanoid CA . We used a custommade inflatable upper body made of nylon fabric with a Double 2 telepresence robot (Double Robotics, Inc.). No-
充气 CA 的核心理念很简单：一个灵活、空的充气机器人作为物理化身。虽然刚性和双臂机器人的工作性能优越，但它们笨重、昂贵，而且接触危险。此外，它们还缺乏虚拟化身的灵活性。拟议的充气 CA 重量轻、价格低，因此很容易复制和定制，类似于数字化身。此外，充气 CA 在放气后可以非常紧凑，即使它们不如数字数据紧凑。在拟议的原型中，远程操作和移动系统是从现有的远程呈现机器人（图 1）改装而来的。人形充气上半身是作为简单网真机器人的可穿戴设备添加的，以构成人形 CA。我们使用尼龙织物制成的定制充气上半身，搭配双人 2 远程呈现机器人（Double Robotics 公司）。无

Fig. 2. Tendon wire mechanism of the inflatable CA. A: an example of wire arrangement for three actuators, and possible anchor points for other deformations. B: position of the eight optical markers for motion capture. C: position of the electrodes for the tactile sensing.
图 2.充气 CA 的腱线机制。A：三个致动器的钢丝布置示例，以及其他变形的可能锚点。B：用于运动捕捉的八个光学标记的位置。C：用于触觉感应的电极位置。
tably, the newer Double 3 robot comprises floor depth sensors and ultrasonic distance sensors below the display that an inflatable cannot cover. To enable the telepresence robot's camera and display, a transparent film window is placed on the front of the chest.
值得注意的是，较新的 Double 3 机器人在显示屏下方装有地面深度传感器和超声波距离传感器，而充气装置无法覆盖这些传感器。为了使远程呈现机器人的摄像头和显示屏能够正常工作，在胸前放置了一个透明薄膜窗口。

An inflatable is a deformable continuum membrane structure supported by inner pressure. As inflatable robots are made of flexible membranes, unlike robots based on a serial-link mechanism, they are fully non-articulated and deformable. Inflatable joints, such as elbows and shoulders, are designed to be relatively malleable to local deformations. Ordinary rigid-body robots have a fixed number of DoFs and actuators determined at the design stage. In contrast, an inflatable CA can increase or decrease the variation in deformation by adding or subtracting actuators on the same body.
充气机器人是由内压支撑的可变形连续膜结构。由于充气机器人是由柔性膜构成的，与基于串联机构的机器人不同，它们是完全非关节化和可变形的。充气关节（如肘部和肩部）的设计对局部变形具有相对的延展性。普通的刚体机器人在设计阶段就已确定了固定数量的 DoF 和致动器。相比之下，充气式 CA 可以通过在同一躯体上增加或减少致动器来增加或减少变形的变化。
We term this concept the variable active DoF. The number of actuators affects the cost of building a robot. It would be beneficial if various deformations could be determined on demand because the types of gestures required by an avatar robot vary depending on the application. For example, assigning the actuators to the neck or arms resulted in variations in the robot's ability for the same number of actuators.
我们将这一概念称为可变主动力系数。执行器的数量会影响机器人的制造成本。如果能按需确定各种变形，那将大有裨益，因为化身机器人所需的手势类型因应用而异。例如，将致动器分配到颈部或手臂上会导致机器人的能力在致动器数量相同的情况下发生变化。

Typical anchor points for human-like movements are shown in Fig. 2A. The actuator used here was an inexpen-
类似人体运动的典型锚点如图 2A 所示。这里使用的致动器是一个不成熟的

Wheeled Inflatable CA 轮式充气 CA
Fig. 3. Schematic diagram of the control system.
图 3.控制系统示意图。
sive RC servo motor (DS3218, Miuzei) with a stall torque of and a speed of when 6.8 V is applied. A swing arm is attached to the motor's output shaft to pull the wire.
在 6.8 V 电压下，RC 伺服电机（DS3218，Miuzei）的失速扭矩为 ，速度为 。电机的输出轴上装有一个摆臂，用于拉动电线。

Capacitive tactile sensors were used to detect the contacts. Inflatable robots do not have a skeleton but are composed of skin-like membranes, enabling the attachment of sensors to the skin. We used a PET film with a flexible and transparent conductive polymer layer (ADFC3, Bit Trade One, Ltd.) as the sensor electrode. Fig. 2C shows the configuration of the electrodes. The capacitance of the electrodes increased with the proximity of the conductive objects, including humans, and a microcontroller board (Arduino Mega) measured the capacitance. The measured value was the count of the time elapsed for the electrode capacitance to charge to a certain voltage. The electrodes were 100,85 , and 0.1 mm in length, width, and thickness, respectively. The electrodes were transparent; therefore, they did not block the display or camera when attached to a clear vinyl window on the front of the inflatable body.
电容式触觉传感器用于检测接触。充气机器人没有骨架，而是由类似皮肤的薄膜组成，因此可以将传感器附着在皮肤上。我们使用带有柔性透明导电聚合物层（ADFC3，Bit Trade One, Ltd.）的 PET 薄膜作为传感器电极。图 2C 显示了电极的构造。电极的电容随着导电物体（包括人类）的靠近而增大，微控制器板（Arduino Mega）测量电容。测量值是电极电容充电到一定电压的时间计数。电极的长度、宽度和厚度分别为 100 毫米、85 毫米和 0.1 毫米。电极是透明的；因此，当电极连接到充气躯体前部的透明乙烯基窗口时，不会遮挡显示屏或摄像头。

An inflatable CA has two controllers. The first is a computer that controls a telepresence robot as a mobile platform. The second is an Arduino microcontroller board that controls the tendon wire mechanism. The user controls the mobile platform through Wi-Fi and Arduino via Bluetooth communication using a PC or smartphone.
充气 CA 有两个控制器。第一个控制器是一台计算机，用于控制作为移动平台的远程呈现机器人。第二个是 Arduino 微控制器板，用于控制腱线机制。用户使用个人电脑或智能手机，通过 Wi-Fi 和 Arduino 的蓝牙通信控制移动平台。
Fig. 3 shows a schematic of the robot's control system.
图 3 显示了机器人控制系统的示意图。

We used three actuators to measure the basic properties. The wires were attached to the anchor points on the left and right elbow joints and throat (Fig. 2A). To pull
我们使用了三个致动器来测量基本特性。电线连接在左右肘关节和喉咙的锚点上（图 2A）。拉动

Fig. 4. Basic movement patterns. Left: nodding. Right: reaching out to shake hands.
图 4.基本运动模式。左：点头。右：伸手握手。
the elbows upward, a plate with a hole in the center was attached to both shoulders as a wire guide. As previously mentioned, an inflatable body has infinite anchor points for deformation.
在肘部向上的位置，在两个肩部连接一块中间有孔的板作为导线。如前所述，充气躯体具有无限的变形锚点。
We measured the robot motion using a motion capture system (OptiTrack Prime 13W, NaturalPoint Inc.) that can measure the positions of optical markers attached to the robot. In this experiment, seven motion capture cameras were used to capture the robot from its surroundings. The temporal resolution of the motion capture was 60 Hz . Markers were attached to the robot at eight locations: the forehead, right shoulder, left shoulder, right elbow, left elbow, right-hand tip, left-hand tip, and feet (Fig. 2B). The motion capture markers were spheres with a diameter of 6.4 mm and were fixed to the marker base. The marker base was attached to the inflatable surface using doublesided tape.
我们使用运动捕捉系统（OptiTrack Prime 13W，NaturalPoint 公司）测量机器人的运动，该系统可以测量机器人上光学标记的位置。在本实验中，我们使用了七台运动捕捉摄像机从周围环境中捕捉机器人的运动。运动捕捉的时间分辨率为 60 Hz。在机器人的前额、右肩、左肩、右肘、左肘、右手尖、左手尖和脚这八个位置都贴有标记（图 2B）。运动捕捉标记是直径为 6.4 毫米的球体，固定在标记基座上。标记基座用双面胶固定在充气表面上。

The prototype inflatable body could perform various movements, including bending the neck and trunk back and forth to the left and right, raising the arms, and bending the elbows. Some combinations of these movements can be used for non-verbal communication, such as approving nodding or greeting bows. In inflatable structures, it is difficult to construct a 3-DoF joint that includes twisting, as observed in the human shoulder. However, it was relatively easy to construct a 1-DoF bending joint. Here, we demonstrated basic movements, elbow flexion, and neck incline, which were distinguishable from other deformations (Fig. 4). The neck tilt was approximately from the vertical, and the elbow flexion was approximately up to from the extended state.
充气躯体原型可以做出各种动作，包括左右前后弯曲颈部和躯干、抬起手臂和弯曲肘部。这些动作的某些组合可用于非语言交流，如赞同点头或问候鞠躬。在充气结构中，很难构造出包括扭转在内的 3-DoF 关节，就像在人类肩部观察到的那样。然而，构建 1-DoF 弯曲关节却相对容易。在这里，我们演示了基本动作、肘关节屈曲和颈部倾斜，它们与其他变形是有区别的（图 4）。颈部倾斜与垂直方向的距离约为 ，肘部弯曲与伸展状态的距离约为 。

The tension of the wires placed inside the inflatable body can deform areas other than the target joint. The previously proposed behavior of the inflatable joints is static [23]. The recovery force against the deformation of an inflatable structure is elastic, as determined by internal pressure and geometry. Here, we proposed a method
放置在充气体内部的导线的张力会使目标关节以外的区域变形。之前提出的充气关节行为是静态的[23]。针对充气结构变形的恢复力是弹性的，由内部压力和几何形状决定。在此，我们提出了一种方法

Fig. 5. The output sinusoidal signal and the transition of the sinusoidal frequency.
图 5 输出正弦信号和正弦频率的过渡输出正弦信号和正弦频率的变化。

Fig. 6. Vertical head sway during nodding movements.
图 6.点头动作时头部的垂直摇摆。
to adjust the movement that exploits the experimental fact that the deformation amplitude varies with frequency in oscillatory movements. The amount of arm swing and torso sway can be adjusted by utilizing or avoiding the resonance points of the inflatable joints.
利用振荡运动中变形幅度随频率变化这一实验事实，对运动进行调节。通过利用或避开充气关节的共振点，可以调整手臂摆动和躯干摇摆的幅度。

In this experiment, the robot performed two movements: (1) an arm-swinging movement and (2) a nodding movement. The arm swing movement was an oscillatory movement in which the left and right elbow joints flexed and extended alternately. The nodding motion involved periodic pulling of the front part of the neck. The frequency of the movements was increased from 0.1 to 1.5 Hz , which was proportional to the elapsed time. The motor command for the servomotors was linear chirp: a sinusoidal wave that increased in frequency. The following equation generates the chirp signal:
在这次实验中，机器人做了两个动作：(1) 摆臂动作和 (2) 点头动作。摆臂运动是左右肘关节交替屈伸的摆动运动。点头动作包括周期性地牵拉颈部前部。动作的频率从 0.1 赫兹增加到 1.5 赫兹，与所用时间成正比。伺服电机的电机指令为线性啁啾波：频率增加的正弦波。啁啾信号由以下公式产生：

represent the frequency, elapsed time from the start of the motion, angle command to the motor, maximum angle, and minimum angle, respectively.
分别代表频率、从运动开始的经过时间、给电机的角度指令、最大角度和最小角度。

Figure 5 shows the output linear chirp command and frequency transition. The frequency of the output sine wave starts at an initial frequency of 0.1 Hz and increases by .
图 5 显示了输出线性啁啾指令和频率转换。输出正弦波的频率 从 0.1 Hz 的初始频率开始，以 的速度递增。

Figure 6 shows the results of the nodding motion. Because the head had a small mass and was located at the center of the torso, the nodding motion was unlikely to cause deformation in other body parts. Movements at low frequencies, such as 0.1 Hz , indicated quasi-static motion. When the frequency was approximately 0.4 Hz , both the forward head inclinations in the backward bending were maximum. The return motion depended on neck elasticity because the nodding motion is a single-wire movement. This implied the possibility of movements of the wire slack depending on the phase and swing of the head motion and motor arm, respectively. The amplitude seemed to decrease as the head and motor motions were close to opposite phases at frequencies around 0.6 Hz . The amplitude, which increased again around 0.8 Hz , decreased with increasing frequency. The results show that complex
图 6 显示了点头运动的结果。由于头部质量较小，且位于躯干中心，点头运动不太可能引起身体其他部位的变形。低频运动，如 0.1 赫兹，表示准静态运动。当频率约为 0.4 赫兹时，向后弯曲时头部的前倾都达到最大值。返回运动取决于颈部弹性，因为点头运动是单线运动。这意味着钢丝松弛的运动可能分别取决于头部运动和运动臂的相位和摆动。在 0.6 赫兹左右的频率下，当头部运动和运动臂运动的相位接近相反时，振幅似乎会减小。振幅在 0.8 赫兹左右再次增大，并随着频率的增加而减小。结果表明

Fig. 7. Comparison of the horizontal swaying of the head and vertical swaying of the hand tips during alternating arm swings.
图 7交替挥臂时头部水平摇摆和手尖垂直摇摆的对比。
amplitude shapes appeared in joints that were not antagonistically driven, where the return motion was uncontrollable.
振幅形状出现在非拮抗驱动的关节中，在这些关节中，返回运动是不可控制的。

Figure 7 shows the results of the arm swing. Only the left-arm hand position is shown because the left and right arms perform the same movement in turns. When the wire was pulled to flex the elbow, tension produced a pulling force on the shoulder. This caused the torso to sway from side to side. Unlike the nodding motion, this is an example of strong interference between arm and trunk motions. The forehead marker position was used
图 7 显示了手臂摆动的结果。只显示了左臂的手部位置，因为左臂和右臂在转弯时的动作相同。当拉线使肘部弯曲时，张力会对肩部产生拉力。这导致躯干左右摇摆。与点头动作不同，这是手臂和躯干动作之间强烈干扰的一个例子。使用前额标记位置

Fig. 8. The measurement of physical interactions with human subjects. From left to right, hug, handshake, and shoulder tap motions.
图 8.测量与人类受试者的身体互动。从左到右依次为拥抱、握手和拍肩动作。
to observe the swaying of the trunk. The input command is a motion with the same amplitude and frequency. We found that the amplitude of the torso sway varied with the frequency. This indicated that the behavior of an inflatable as an air spring provides unique dynamic characteristics. Movements at low frequencies (approximately 0.1 Hz ) indicated quasi-static motion. The amplitude of the oscillations continued to increase until approximately 0.9 Hz , with significant decreases at higher frequencies. The results showed that frequencies around 0.9 Hz are suitable for efficiently obtaining elbow flexion and leftright movement of the torso with a single actuator. However, elbow flexion should be performed rapidly to avoid swaying the body.
来观察躯干的摇摆。输入指令是振幅和频率相同的运动。我们发现，躯干摇摆的幅度随频率而变化。这表明作为空气弹簧的充气躯干具有独特的动态特性。低频运动（约 0.1 赫兹）表明是准静态运动。振荡幅度持续增加，直到大约 0.9 赫兹，频率越高，幅度越小。结果表明，0.9 赫兹左右的频率适合使用单个致动器有效地实现肘部屈曲和躯干左右运动。不过，屈肘动作应快速进行，以避免身体晃动。

We performed preliminary experiments to detect the physical contact between the human subjects and the robot using the implemented tactile sensors. We selected hugs, handshakes, and shoulder taps as typical nonverbal physical interactions with different body parts. However, this experiment was a preliminary stage of physical interaction. The robot did not facilitate contact, approach the human, or respond to contact with the human. Three adult participants were included in the study. They were instructed to perform a hug, handshake, and shoulder tap on the inflatable humanoid CA at specific intervals. The number of times each movement was specified, but not the duration or intensity. The subjects were instructed to perform one hug, two handshakes, and three shoulder taps. The CA mobile platform was maintained in a standing position and was not moved.
我们进行了初步实验，利用已实现的触觉传感器检测人类受试者与机器人之间的身体接触。我们选择了拥抱、握手和拍肩膀作为不同身体部位的典型非语言物理交互。不过，这次实验只是身体互动的初级阶段。机器人并没有促进接触、接近人类或对人类的接触做出反应。三名成年参与者参与了这项研究。他们被要求在特定时间间隔内对充气人形 CA 进行拥抱、握手和拍肩等动作。每个动作的次数都有规定，但没有规定持续时间或强度。受试者被要求进行一次拥抱、两次握手和三次拍肩。CA 移动平台保持站立姿势，不得移动。

The measurement results of the physical contacts using tactile sensors are shown in Fig. 8. The electrodes were not placed on the entire inflatable membrane but in locations corresponding to each contact. The vertical movement of the arm during the handshake did not significantly
使用触觉传感器对物理接触的测量结果如图 8 所示。电极没有放置在整个充气膜上，而是放置在每次接触的相应位置。在握手过程中，手臂的垂直运动并没有明显影响触觉传感器的测量结果。

Fig. 9. The inflatable robots in the hybrid events.
图 9.混合赛事中的充气机器人。
affect the sensor values. Large individual differences were observed in the contact duration and the sensor values amplitude. In addition, the experimental results showed that detecting fast contact transitions during shoulder tapping is possible. This experiment was a one-time preliminary experiment and did not verify whether it was possible to classify gesture types or subjects based on sensor data.
影响传感器数值。在接触持续时间和传感器数值振幅方面，观察到了巨大的个体差异。此外，实验结果表明，检测拍肩过程中的快速接触转换是可能的。本实验只是一次性的初步实验，并未验证是否可以根据传感器数据对手势类型或受试者进行分类。

The proposed inflatable CA was used at a symposium in the fall of 2021. The symposium was a hybrid event with keynote speeches and panel discussions conducted in a physical lecture hall and streamed online. During the panel discussion, the robot moved freely before the six panelists as the only audience (and on behalf of the online audience) and responded with gestures. To avoid hiding the seated panelists, the inflatable CA was not on the same stage but on a slightly lower front row as a moving space (Fig. 9). Although the moderator could have technically
在 2021 年秋季的一次研讨会上使用了拟议的充气 CA。研讨会是一个混合活动，包括在实体演讲厅进行的主题演讲和小组讨论，以及在线流媒体。在小组讨论期间，机器人作为唯一的观众（并代表在线观众）在六位小组成员面前自由移动，并用手势做出回应。为了避免遮挡坐着的小组成员，充气 CA 并不在同一个舞台上，而是在稍低的前排作为移动空间（图 9）。虽然主持人在技术上
allowed speaking through the robot, it avoided simplifying the system of streaming the voice to the physical lecture hall and online video.
通过机器人发言，可以避免简化将语音流传输到实体讲堂和在线视频的系统。

We performed a timely playback of predefined gestures, including elbow bending, nodding at different speeds, and torso swaying obtained by alternately bending the left and right elbows. For safety reasons, the two operators sat in a hall. They were tasked with moving the robot and activating their gestures. One operator chose gestures from a list of nine, and the computer sent the actual commands to the CA. The other operator controlled the mobile platform from the computer using the user interface provided by Double Robotics, Inc. The mobile platform could move forward and backward, turn left and right, and then stop in place.
我们及时播放了预先设定的手势，包括弯肘、以不同速度点头以及通过左右肘部交替弯曲获得的躯干摇摆。出于安全考虑，两名操作员坐在一个大厅里。他们的任务是移动机器人并启动手势。其中一名操作员从九个手势列表中选择手势，计算机将实际指令发送给 CA。另一名操作员则使用 Double Robotics 公司提供的用户界面从电脑上控制移动平台。移动平台可以前后移动、左右转弯，然后停在原地。

The moderator and panelists were prevented from leaving their seats and separated by transparent partitions to prevent infection. Only the robot could approach people without a mask on the stage. The event involved 600 online participants.
主持人和小组成员不能离开座位，并用透明隔板隔开，以防感染。只有机器人可以接近台上没有戴面具的人。这次活动有 600 名在线参与者。

We hypothesized that owing to the inherent lightness and softness of the proposed inflatable humanoid CA, psychological safety will enable HRI experiments that are impossible with conventional rigid robots. However, this proposal was limited to the implementation of a platform. More elaborate subject experiments must be conducted by considering individual interaction scenarios to investigate the various characteristics of inflatable CA.
我们假设，由于拟议中的充气仿人 CA 本身轻盈柔软，心理安全将使传统刚性机器人无法进行的 HRI 实验成为可能。然而，这一建议仅限于一个平台的实施。必须通过考虑单个交互场景来进行更详细的主体实验，以研究充气 CA 的各种特性。

The current method of activating fixed gestures and remote control in a robot through direct observation by an operator is a clear disadvantage. The challenges are how to directly reflect the operator's movements on the inflatable CA and improve the inflatable joints' DOF and range of motion for this purpose. Furthermore, communication delays and the stable operation of the robot will be an issue for remote operation from a considerable distance.
目前通过操作员的直接观察来激活机器人的固定手势和远程控制的方法存在明显的缺点。如何将操作员的动作直接反映在充气 CA 上，并为此改进充气关节的多自由度和运动范围，是目前面临的挑战。此外，通信延迟和机器人的稳定运行也是远距离远程操作的一个问题。

Detailed measurement of the soft deformation of nonrigid inflatable structures is another challenge. Because inflatables are hollow in structure, there is no axis of rotation at the joints, which makes it difficult to measure joint angles. Optical motion capture is an effective tool, but posture changes cannot be measured using a small number of markers.
对非刚性充气结构的软变形进行详细测量是另一项挑战。由于充气式结构是空心的，关节处没有旋转轴，因此很难测量关节角度。光学运动捕捉是一种有效的工具，但使用少量标记无法测量姿势变化。

The modeling challenges should be mentioned. The behavior of the membrane structures, including wrinkles, is complex. Therefore, a suitable computational model of inflatables is required to design the shape of the inflatables, which currently relies on a manual design. This study investigated the dynamic characteristics of inflatables regarding relatively simple oscillations. Therefore, a mathematical model that can explain the observed behavior is necessary.
值得一提的是建模方面的挑战。膜结构（包括皱褶）的行为非常复杂。因此，需要一个合适的充气艇计算模型来设计充气艇的形状，而目前只能依靠人工设计。本研究调查了充气艇相对简单振荡的动态特性。因此，有必要建立一个能够解释观察到的行为的数学模型。

Humanoid robots can act as communication channels that keep humans in sync with the cyber world. We focused on a lightweight and inexpensive inflatable structure to realize a scalable avatar robot. The proposed inflatable humanoid CA was achieved by adding humanlike physicality to an easy-to-handle telepresence robot. An inflatable CA can have variable active degrees of freedom. A soft inflatable structure allowed the same body to perform a different set of gestures depending on the actuator arrangement. In addition, we implemented tactile sensing in the proposed humanoid CA. The experimental results with different frequency inputs show that the neck joint had two peaks in amplitude. Furthermore, we found coupling of the elbow motion and torso sway with a resonance point. Finally, we conducted experiments to detect physical interactions such as hugs, handshakes, and shoulder taps. The proposed robot was operated in a hybrid event. Future challenges include modeling the dynamic characteristics of inflatables, implementing different gestures, and performing experiments with rich interaction scenarios.
仿人机器人可以作为沟通渠道，让人类与网络世界保持同步。我们重点研究了一种轻便、廉价的充气结构，以实现可扩展的化身机器人。我们提出的可充气仿人 CA 是通过在易于操作的远程呈现机器人上添加类似人类的物理特性来实现的。充气 CA 可以具有可变的活动自由度。柔软的充气结构允许同一躯体根据致动器的布置做出不同的手势。此外，我们还在拟人 CA 中实现了触觉传感。不同频率输入的实验结果显示，颈部关节有两个振幅峰值。此外，我们还发现肘部运动和躯干摇摆之间存在耦合共振点。最后，我们还进行了实验，以检测拥抱、握手和拍肩等身体互动。拟议的机器人在混合活动中进行了操作。未来的挑战包括为充气娃娃的动态特性建模、实现不同的手势以及进行丰富的交互场景实验。

This work was supported by the JST Moonshot R&D Grant Number JPMJMS2013. We appreciate Babot Inc. (Tokyo, Japan) for the manufacturing of the inflatables.
本研究得到了 JST Moonshot R&D 补助金（编号：JPMJMS2013）的支持。我们感谢 Babot 公司（日本东京）制造了充气娃娃。

[1] K. Hoshino, N. Igo, M. Tomida, and H. Kotani, "Teleoperating System for Manipulating a Moon Exploring Robot on the Earth," Int. J. Automation Technol., Vol.11, No.3, pp. 433-441, 2017.
[1] K. Hoshino, N. Igo, M. Tomida, and H. Kotani, "Teleoperating System for Manipulating a Moon Exploring Robot on the Earth," Int. J. Automation Techniques, Vol.自动化技术》，第 11 卷第 3 期，第 433-441 页，2017 年。
[2] H. Tezuka, N. Katafuchi, Y. Nakamura, T. Machino, Y. Nanjo, S. Iwaki, and K. Shimokura, "Robot Platform Architecture for Information Sharing and Collaboration Among Multiple Networked Robots," J. Robot. Mechatron., Vol.18, No.3, pp.325-332, 2006.
[2] H. Tezuka, N. Katafuchi, Y. Nakamura, T. Machino, Y. Nanjo, S. Iwaki, and K. Shimokura, "Robot Platform Architecture for Information Sharing and Collaboration Among Multiple Networked Robots," J. Robot.机电一体化》，第 18 卷第 3 期，第 325-332 页，2006 年。
[3] H. O. Khadri, "University academics' perceptions regarding the future use of telepresence robots to enhance virtual transnational education: an exploratory investigation in a developing country," Smart Learning Environments, Vol.8, No.1, 28, 2021.
[3] H. O. Khadri，"大学学者对未来使用远程呈现机器人加强虚拟跨国教育的看法：在一个发展中国家进行的探索性调查"，《智能学习环境》，第 8 卷第 1 期，28，2021 年。
[4] S. Tachi, "Telexistence: Enabling Humans to Be Virtually Ubiquitous," IEEE Computer Graphics and Applications, Vol.36, No.1, pp. 8-14, 2016.
[4] S. Tachi，"Telexistence：Enabling Humans to Be Virtually Ubiquitous，" IEEE Computer Graphics and Applications，Vol.36，No.1，pp.8-14，2016.
[5] C. L. Fernando, M. Furukawa, T. Kurogi, S. Kamuro, K. Sato, K. Minamizawa, and S. Tachi, "Design of TELESAR V for transferring bodily consciousness in telexistence," IEEE/RSJ Int. Conf. on Intelligent Robots and Systems (IROS), pp. 5112-5118, 2012.
[5] C. L. Fernando、M. Furukawa、T. Kurogi、S. Kamuro、K. Sato、K. Minamizawa 和 S. Tachi，"TELESAR V 的设计用于在 Telexistence 中传递身体意识"，IEEE/RSJ Int.Conf. on Intelligent Robots and Systems (IROS), pp.
[6] J. Krüger, G. Schreck, and D. Surdilovic, "Dual arm robot for flexible and cooperative assembly," CIRP Annals, Vol.60, No.1, pp. 5-8, 2011.
[6] J. Krüger、G. Schreck 和 D. Surdilovic，"用于灵活协同装配的双臂机器人"，《CIRP 年鉴》，第 60 卷，第 1 期，第 5-8 页，2011 年。
[7] B. Wang, H. Zhou, G. Yang, X. Li, and H. Yang, "Human Digital Twin (HDT) Driven Human-Cyber-Physical Systems: Key Technologies and Applications," Chinese J. of Mechanical Engineering, Vol.35, No.1, 11, 2022.
[7] B. Wang, H. Zhou, G. Yang, X. Li, and H. Yang, "Human Digital Twin (HDT) Driven Human-Cyber-Physical Systems：关键技术与应用》，《机械工程学报》，第 35 卷第 1 期，2022 年 11 月。
[8] T. Tamei, T. Matsubara, A. Rai, and T. Shibata, "Reinforcement learning of clothing assistance with a dual-arm robot," IEEE-RAS Int. Conf. on Humanoid Robots (Humanoids), pp. 733-738, 2011.
[8] T. Tamei, T. Matsubara, A. Rai, and T. Shibata, "Reinforcement learning of clothing assistance with a dual-arm robot," IEEE-RAS Int. Conf. on Humanoid Robots (Humanoids), pp.Conf. on Humanoid Robots (Humanoids), pp.
[9] T. Inamura, Y. Mizuchi, and H. Yamada, "VR platform enabling crowdsourcing of embodied HRI experiments - case study of online robot competition," Vol.35, No.11, pp. 697-703, 2021.
[9] T. Inamura, Y. Mizuchi, and H. Yamada, "VR platform enabling crowdsourcing of embodied HRI experiments - case study of online robot competition," Vol.35, No.11, pp.
[10] A. E. Block, H. Seifi, O. Hilliges, R. Gassert, and K. J. Kuchenbecker, "In the Arms of a Robot: Designing Autonomous Hugging Robots with Intra-Hug Gestures," J. Human-Robot Interaction, pp. 37-42, 2022.
[10] A. E. Block, H. Seifi, O. Hilliges, R. Gassert, and K. J. Kuchenbecker, "In the Arms of a Robot：Designing Autonomous Hugging Robots with Intra-Hug Gestures," J. Human-Robot Interaction, pp.
[11] A. E. Block and K. J. Kuchenbecker, "Softness, Warmth, and Responsiveness Improve Robot Hugs," Int. J. of Social Robotics, Vol.11, No.1, pp. 49-64, 2019.
[11] A. E. Block and K. J. Kuchenbecker, "Softness, Warmth, and Responsiveness Improve Robot Hugs," Int.J. of Social Robotics, Vol.11, No.1, pp.
[12] S. Higuchi and H. Oku, "Wide angular range dynamic projection mapping method applied to drone-based avatar robot," Advanced Robotics, Vol.35, No.11, pp. 675-684, 2021.
[12] S. Higuchi 和 H. Oku：《应用于无人机化身机器人的广角范围动态投影映射法》，《先进机器人学》第 35 卷第 11 期，第 675-684 页，2021 年。
[13] R. Niiyama, H. Sato, K. Tsujimura, K. Narumi, Y. A. Seong, R. Yamamura, Y. Kakehi, and Y. Kawahara, "Poimo: Portable and Inflatable Mobility Devices Customizable for Personal Physical Characteristics," Annual ACM Symp. on User Interface Software and Technology (UIST), pp. 912-923, 2020.
[13] R. Niiyama、H. Sato、K. Tsujimura、K. Narumi、Y. A. Seong、R. Yamamura、Y. Kakehi 和 Y. Kawahara，"Poimo：可根据个人身体特征定制的便携式充气移动设备"，ACM 用户界面软件与技术（UIST）年度研讨会，第 912-923 页，2020 年。
[14] S. Sanan, M. H. Ornstein, and C. G. Atkeson, "Physical human interaction for an inflatable manipulator," Int. Conf. the IEEE Engineering in Medicine and Biology Society (EMBS), pp. 7401-7404, 2011.
[14] S. Sanan、M. H. Ornstein 和 C. G. Atkeson，"充气机械手的物理人机交互"，Int.Conf. the IEEE Engineering in Medicine and Biology Society (EMBS), pp.
[15] A. Alspach, J. Kim, and K. Yamane, "Design and fabrication of a soft robotic hand and arm system," IEEE Int. Conf. on Soft Robotics (RoboSoft), pp. 369-375, 2018.
[15] A. Alspach, J. Kim, and K. Yamane, "Design and fabrication of a soft robotic hand and arm system," IEEE Int. Conf. on Soft Robotics (RoboSoft), pp.Conf. on Soft Robotics (RoboSoft), pp.
[16] F. Garzotto, M. Gelsomini, and Y. Kinoe, "Puffy: A Mobile Inflatable Interactive Companion for Children with Neurodevelopmental Disorder," Human-Computer Interaction - INTERACT 2017, pp. 467-492, 2017.
[16] F. Garzotto, M. Gelsomini, and Y. Kinoe, "Puffy: A Mobile Inflatable Interactive Companion for Children with Neurodevelopmental Disorder," Human-Computer Interaction - INTERACT 2017, pp.
[17] J. Jaramillo, A. Lin, E. Sung, I. J. H. Richter, and K. Petersen, "Mobile, Inflatable Interface to Support Human Robot Interaction Studies," Int. Conf. on Ubiquitous Robots (UR), pp. 320-325, 2021.
[17] J. Jaramillo、A. Lin、E. Sung、I. J. H. Richter 和 K. Petersen，"支持人机交互研究的移动充气界面"，Int.Ubiquitous Robots (UR), pp.
[18] R. Qi, T. L. Lam, and Y. Xu, "Mechanical design and implementation of a soft inflatable robot arm for safe human-robot interaction," IEEE Int. Conf. on Robotics and Automation (ICRA), pp. 34903495, 2014
[18] R. Qi, T. L. Lam, and Y. Xu, "Mechanical design and implementation of a soft inflatable robot arm for safe human-robot interaction," IEEE Int. Conf. on Robotics and Automation (ICRA), pp.Robotics and Automation (ICRA), pp.
[19] J. M. Bern, K-H. Chang, and S. Coros, "Interactive Design of Animated Plushies," ACM Trans. on Graphics, Vol.36, No.4, pp. 1-11, 2017.
[19] J. M. Bern、K-H.Chang, and S. Coros, "Interactive Design of Animated Plushies," ACM Trans. on Graphics, Vol.36, No.4, pp.
[20] Y. Takase, H. Mitake, Y. Yamashita, and S. Hasegawa, "Motion generation for the stuffed-toy robot," Annual Conf. of the Society of Instrument and Control Engineers of Japan, pp. 213-217, 2013
[20] Y. Takase、H. Mitake、Y. Yamashita 和 S. Hasegawa，"毛绒玩具机器人的运动生成"，日本仪器与控制工程师学会年度会议，第 213-217 页，2013 年。
[21] R. Niiyama, Y. A. Seong, Y. Kawahara, and Y. Kuniyoshi, "BlowerPowered Soft Inflatable Joints for Physical Human-Robot Interaction," Frontiers in Robotics and AI, Vol.8, pp. 1-12, 2021.
[21] R. Niiyama、Y. A. Seong、Y. Kawahara 和 Y. Kuniyoshi，"用于物理人机交互的鼓风机驱动软充气关节"，《机器人学和人工智能前沿》，第 8 卷，第 1-12 页，2021 年。
[22] H. Ishiguro, "The realisation of an avatar-symbiotic society where everyone can perform active roles without constraint," Advanced Robotics, Vol.35, No.11, pp. 650-656, 2021.
[22] H. Ishiguro，"人人都能无拘无束地发挥积极作用的化身共生社会的实现"，《先进机器人技术》，第 35 卷第 11 期，第 650-656 页，2021 年。
[23] Y. A. Seong, R. Niiyama, Y. Kawahara, and Y. Kuniyoshi, "Lowpressure soft inflatable joint driven by inner tendon," 2nd IEEE Int. Conf. on Soft Robotics (RoboSoft), pp. 37-42, 2019.
[23] Y. A. Seong, R. Niiyama, Y. Kawahara, and Y. Kuniyoshi, "Lowpressure soft inflatable joint driven by inner tendon," 2nd IEEE Int. Conf. on Soft Robotics (RoboSoft), pp.软机器人（RoboSoft）会议，第 37-42 页，2019 年。

Ryuma Niiyama 新山龙马

Senior Assistant Professor, Graduate School of Science and Technology, Meiji University
明治大学科学技术研究生院高级助理教授

1-1-1 Higashi-Mita, Tama-ku, Kawasaki-shi, Kanagawa 214-8571, Japan Brief Biographical History:
日本神奈川县川崎市多摩区东三田 1-1-1 号（邮编 214-8571）简历：
2010 Received Ph.D. from The University of Tokyo
2010 获东京大学博士学位
2014- Lecturer, The University of Tokyo
2014- 东京大学讲师
2022- Senior Assistant Professor, Meiji University
2022- 明治大学高级助理教授

Research Fellow, School of Science and Technology, Meiji University
明治大学科学技术学院研究员

1-1-1 Higashi-Mita, Tama-ku, Kawasaki-shi, Kanagawa 214-8571, Japan Brief Biographical History:
日本神奈川县川崎市多摩区东三田 1-1-1 号（邮编 214-8571）简历：
2012-2014 TMT Machinery, Inc.
2012-2014 TMT 机械公司
2019 Received Ph.D. from Tokyo University of Agriculture and Technology
2019 获得东京农工大学博士学位
2019-2022 Research Fellow, The University of Tokyo
2019-2022 东京大学研究员
2022- Research Fellow, Meiji University
2022- 明治大学研究员

Young Ah Seong

0000-0003-1967-8084

Visiting Associate Professor, Hosei University
法政大学客座副教授

2-17-1 Fujimi, Chiyoda-ku, Tokyo 102-8160, Japan
102-8160 日本东京千代田区富士见 2-17-1

2012- Samsung Electronics Co., Ltd.
2012- 三星电子有限公司
2017- Project Researcher, The University of Tokyo
2017- 东京大学，项目研究员
2020- Visiting Associate Professor, Hosei University
2020- 法政大学客座副教授