2024_08_11_345044714115311b2ddbg

Companion Unmanned Aerial Vehicles: A Survey
配套无人驾驶飞行器：调查

Chun Fui Liew and Takehisa Yairi
Chun Fui Liew 和 Takehisa Yairi

Abstract 摘要

Recent technological advancements in smallscale unmanned aerial vehicles (UAVs) have led to the development of companion UAVs. Similar to conventional companion robots, companion UAVs have the potential to assist us in our daily lives and to help alleviating social loneliness issue. In contrast to ground companion robots, companion UAVs have the capability to fly and possess unique interaction characteristics. Our goals in this work are to have a bird's-eye view of the companion UAV works and to identify lessons learned and guidelines for the design of companion UAVs. We tackle two major challenges towards these goals, where we first use a coordinated way to gather topquality human-drone interaction (HDI) papers from three sources, and then propose to use a perceptual map of UAVs to summarize current research efforts in HDI. While being simple, the proposed perceptual map can cover the efforts have been made to realize companion UAVs in a comprehensive manner and lead our discussion coherently. We also discuss patterns we noticed in the literature and some lessons learned throughout the review. In addition, we recommend several areas that are worth exploring and suggest a few guidelines to enhance HDI researches with companion UAVs.
近年来，小型无人驾驶飞行器（UAVs）技术的进步促进了陪伴型无人驾驶飞行器的发展。与传统的陪伴机器人类似，陪伴型无人机有可能在日常生活中为我们提供帮助，并有助于缓解社会孤独问题。与地面陪伴机器人相比，陪伴型无人机具有飞行能力和独特的交互特性。我们在这项工作中的目标是鸟瞰陪伴型无人机的工作，并为陪伴型无人机的设计找出经验教训和指导方针。为了实现这些目标，我们应对了两大挑战：首先，我们采用协调的方式从三个来源收集高质量的人机交互（HDI）论文；然后，我们建议使用无人机感知地图来总结当前的人机交互研究工作。提出的感知图虽然简单，却能全面涵盖为实现陪伴型无人机所做的努力，并能连贯地引导我们的讨论。我们还讨论了我们在文献中注意到的模式，以及在整个审查过程中汲取的一些经验教训。此外，我们还建议了几个值得探索的领域，并提出了一些指导原则，以加强对伴航无人机的人类发展指数研究。

Keywords Unmanned aerial vehicle (UAV), micro aerial vehicle (MAV), aerial robotics, flying robots, drone, companion UAV, social UAV, proximate interaction, collocated interaction, human-robot interaction, human-drone interaction
关键词无人飞行器（UAV）、微型飞行器（MAV）、空中机器人、飞行机器人、无人机、陪伴型无人飞行器、社交型无人飞行器、近距离交互、协同交互、人与机器人交互、人与无人机交互

1 Introduction 1 引言

Companion robots are robots designed to have social interaction and emotional connection with people. For examples, companion robots like Paro (a therapeutic robotic seal for elderly) 112, EmotiRob (a companion robotic bear for children) 106, Aibo (a companion robotic dog) [50], and Jibo (a companion robot in home) 47 could interact socially with people. One potential application of companion robots is that they could be our personal assistance. One could also argue that companion robots are similar to pets and they help to alleviate social isolation or loneliness issues.
陪伴机器人是指设计用于与人进行社交互动和情感交流的机器人。例如，帕罗（老年人治疗机器人海豹）112、EmotiRob（儿童陪伴机器人小熊）106、Aibo（陪伴机器人狗）[50] 和 Jibo（家庭陪伴机器人）47 等陪伴机器人可以与人进行社交互动。陪伴机器人的一个潜在应用是，它们可以成为我们的个人助手。也有人认为，陪伴机器人类似于宠物，有助于缓解社会隔离或孤独问题。

Recently, technological advancements have led to a new class of companion robots-a companion unmanned aerial vehicle (UAV). Compared to conventional companion robots, companion UAVs have some distinctive characteristics-notably their capability to fly (hence their new interaction capability) and their strict design constraints such as safety concern, noise issue, flight time and payload limitation. While there are still many technical difficulties and social design questions, various concepts, design sketches, and prototypes have been proposed to demonstrate the idea of a companion UAV, including a flying jogging companion [36, a flying assistant that could interact with us in daily tasks [111, a human-centered designed drone aiming to fly in a human crowd environment 118, a flying smart agent that could assist users through active physical participation [3], flying "fairies" 27, 74 and flying lampshades 100, 101 that could dance with human on the stage, a flying ball for augmented sports 78, a flying humanoid balloon to accompany children 22 , a companion UAV that can react to human emotions 67, and a moving projector platform for street games 54 .
最近，技术进步催生了一种新型陪伴机器人--陪伴式无人飞行器（UAV）。与传统的陪伴机器人相比，陪伴型无人飞行器具有一些与众不同的特点--尤其是其飞行能力（因此具有新的交互能力）和严格的设计限制，如安全问题、噪音问题、飞行时间和有效载荷限制。尽管还存在许多技术难题和社会设计问题，但人们已经提出了各种概念、设计草图和原型来展示伴飞无人机的理念，其中包括飞行慢跑伴侣[36]、可与我们在日常任务中互动的飞行助手[111]、以人为本设计的无人机，旨在人类人群环境中飞行[118]、可以通过主动身体参与协助用户的飞行智能代理[3]，可以在舞台上与人类共舞的飞行 "仙女 "27、74 和飞行灯罩 100、101，用于增强运动的飞行球 78，陪伴儿童的飞行人形气球 22，可以对人类情绪做出反应的陪伴型无人机 67，以及用于街头游戏的移动投影仪平台 54。

Our goals in this survey are to have a bird's-eye view of the companion UAV works and to identify lessons
我们此次调查的目的是鸟瞰配套无人机的工作情况，并总结经验教训。
learned, guidelines, and best practices for the design of companion UAVs. There are two major challenges towards these goals: (i) to find a coordinated way to identify top-quality HDI works from a huge amount of available literature, and (ii) to find a suitable framework or organizing principle to discuss the vast aspects of existing human-drone interaction (HDI) works.
设计配套无人机的经验、准则和最佳做法。实现这些目标有两大挑战：(i)找到一种协调的方法，从大量可用文献中识别高质量的人机交互作品；(ii)找到一个合适的框架或组织原则，以讨论现有人机交互（HDI）作品的方方面面。

To tackle the first challenge, i.e., to cover top-quality companion UAV works as comprehensive as possible in this review, we use a coordinated way to gather HDI papers from three major sources. First, we systematically identified 1,973 high-quality UAV papers from more than forty-eight thousand general papers that have appeared in the top robotic journals and conferences since 2001. In a nutshell, this identification process involves a few steps consist of automated and manual processes (more details in Section 3). Second, from the identified UAV papers, we tagged the papers with several key topics, studied those related to the topic of HRI and analyzed their references, and continued to track down HDI-related papers. Third, we included HDI papers recommended by reviewers during our past journal submission.
为了应对第一个挑战，即在本综述中尽可能全面地涵盖高质量的伴飞无人机作品，我们采用协调的方式从三大来源收集人类发展指数论文。首先，我们从 2001 年以来发表在顶级机器人期刊和会议上的四万八千多篇普通论文中，系统地识别出 1973 篇高质量的无人机论文。简而言之，这一识别过程包括几个步骤，由自动和人工流程组成（详情见第 3 节）。其次，从识别出的无人机论文中，我们用几个关键主题对论文进行标记，研究那些与人类资源创新主题相关的论文并分析其参考文献，然后继续追踪与人类资源创新相关的论文。第三，我们收录了以往期刊投稿时审稿人推荐的人类发展指数论文。

To tackle the second challenge, i.e., to find a suitable framework to discuss the vast aspects of existing HDI works, we propose to use a perceptual map of UAVs (more details in Section 4) as a high-level framework to organize current research efforts in HDI. In the proposed perceptual map, we categorize UAVs based on the degree of autonomy and the degree of sociability. This basic categorization leads to four distinct categories, namely remotely-controlled UAV, autonomous UAV, social UAV, and companion UAV. Looking at the research and development of companion UAVs with this perceptual map, we can find two main direction of on-going efforts. Moreover, we find this perceptual map easy to understand and lead our discussion coherently.
为了应对第二个挑战，即找到一个合适的框架来讨论现有人类发展倡议工作的方方面面，我们建议使用无人机感知图（更多详情见第4节）作为一个高层次框架来组织当前的人类发展倡议研究工作。在建议的感知地图中，我们根据无人机的自主程度和社交程度对其进行分类。这种基本分类方法可分为四个不同的类别，即遥控无人机、自主无人机、社交无人机和陪伴无人机。根据这一感知图谱来审视陪伴型无人机的研究与开发，我们可以发现两个主要的努力方向。此外，我们还发现该感知图易于理解，并能连贯地引导我们的讨论。

This work emphasizes on the proximate interaction between a human and a companion UAV in the HDI field. Note that proximate interaction is also called collocated interaction in some literature [38, 115]. In the following sections, we first briefly explain the definition of a UAV and different type of UAVs (Section 2) in order to facilitate the discussion in this work. Next, we describe methodology we used to identify top-quality UAV papers from the literature (Section 3). Then, we discuss the perceptual map of UAVs (Section (4), followed by discussion on research efforts in realizing companion UAVs from the engineering (Section 5) and sociability (Section 6) perspectives. In Section 7 and Section 8 , we discussion several observation and lessons learned, along with guidelines and recommendations for realizing a companion UAV. Section 9 draw conclusion about future research directions for companion UAVs.
这项工作的重点是人类与伴飞无人机在人类发展指数领域的近距离交互。请注意，在一些文献中，近距离交互也被称为同地交互 [38, 115]。在下面的章节中，我们首先简要解释无人机的定义和不同类型的无人机（第 2 节），以便于本作品的讨论。接下来，我们将介绍从文献中识别高质量无人机论文的方法（第 3 节）。然后，我们讨论了无人机的感知图谱（第 4 节），接着从工程学（第 5 节）和社交性（第 6 节）的角度讨论了实现陪伴型无人机的研究工作。在第 7 节和第 8 节中，我们讨论了一些观察结果和经验教训，以及实现陪伴型无人机的指导方针和建议。第 9 节总结了陪伴型无人机的未来研究方向。

2 UAV Background 2 无人驾驶飞行器背景

We first explain the UAV definition and introduce some common UAV types to facilitate the following discussion. We recommend the handbook of UAVs [109 if readers are interested in the more technical details of UAV.
我们首先解释无人飞行器的定义，并介绍一些常见的无人飞行器类型，以方便下面的讨论。如果读者对无人飞行器的更多技术细节感兴趣，我们推荐阅读《无人飞行器手册》[109]。

2.1 UAV Definition 2.1 无人机定义

UAVs, commonly known as drones, are aircraft that can perform flight missions without a human pilot onboard 30. In general, UAVs can be viewed as flying robots. The UAV's degree of autonomy varies but often modern UAVs are able to hover stably at a point in 3D space. UAVs with a higher degree of autonomy offer more functions like automatic take-off and landing, path planning, and obstacle avoidance. In the literature, UAVs have several other names such as micro aerial vehicle (MAV), unmanned aerial system (UAS), vertical take-off and landing aircraft (VTOL), multicopter, rotorcraft, and aerial robot. In this work, we will use the "UAV" and "drone" terms interchangeably.
无人驾驶飞行器通常被称为无人机，是一种可以在没有人类飞行员的情况下执行飞行任务的飞行器 30。一般来说，无人飞行器可被视为飞行机器人。无人机的自主程度各不相同，但现代无人机通常能够在三维空间的某一点稳定悬停。自主程度较高的无人机可提供更多的功能，如自动起降、路径规划和避障。在文献中，无人机还有其他一些名称，如微型飞行器（MAV）、无人机系统（UAS）、垂直起降飞机（VTOL）、多旋翼机、旋翼机和空中机器人。在本文中，我们将交替使用 "UAV "和 "无人机 "这两个术语。

2.2 UAV Types 2.2 无人机类型

Conventionally, UAVs can be classified into fixed-wing, multirotor, blimp, or balloon types based on their flying principle. In the end of this review, one could observe that most UAV prototypes we described in this work are multirotor UAVs. We speculate this is due to the availability of multirotor type UAVs in the market. In Section 7, we will have a more rigorous discussion arguing that the blimp or balloon UAVs could be a better form for companion UAVs.
按照飞行原理，无人机通常可分为固定翼、多旋翼、飞艇或气球等类型。在本综述的最后，我们可以发现，我们在本作品中描述的大多数无人机原型都是多旋翼无人机。我们推测，这是由于市场上存在多旋翼无人机。在第 7 节中，我们将进行更严谨的讨论，认为飞艇或气球无人机可能是更好的伴飞无人机形式。

It is worth noting that Floreano & Wood 33 have classified UAVs based on the flight time and UAV mass (a simplified plot is shown in Fig. 17). In general, flappingwing UAVs are small and have a short flight time. Blimp/balloon UAVs are lightweight and have a longer flight time. Rotor-type and fixed-wing UAVs are usually heavier. In Section 7, we will have more discussion about the safety and noise issues of different types of UAVs.
值得注意的是，Floreano & Wood 33 根据飞行时间和无人机质量对无人机进行了分类（简化图见图 17）。一般来说，拍翼无人机体积小，飞行时间短。飞艇/气球无人机重量轻，飞行时间较长。旋翼式和固定翼无人机通常较重。在第 7 节中，我们将进一步讨论不同类型无人机的安全和噪音问题。

3 UAV Paper Identification Process
3 无人机纸张识别过程

The UAV papers identification process involves three major steps. We first used a script to automatically collect more than forty-eight thousand instances of title and abstract from fourteen top journal/conference web pages since 2001 (the seven journals include IEEE Transactions on Robotics (TRO), IEEE/ASME Transactions on Mechatronics (TME), The International Journal
无人机论文识别过程包括三个主要步骤。首先，我们使用一个脚本，从 2001 年以来的 14 个顶级期刊/会议网页中自动收集了四万八千多个标题和摘要实例（这 7 个期刊包括：《IEEE 机器人学学报》（TRO）、《IEEE/ASME 机电学报》（TME）、《The International Journal》、《The International Journal》、《The International Journal》、《The International Journal》、《The International Journal》、《The International Journal》、《The International Journal》、《The International Journal》、《The International Journal》、《The International Journal》）。

Fig. 1 UAV types based on flight time and UAV mass (inspired by Floreano & Wood 33).
图 1 基于飞行时间和无人机质量的无人机类型（灵感来自 Floreano & Wood 33）。

Table 135 keywords used to search drone papers systematically from the collected titles and abstracts.
表 135 从收集到的标题和摘要中系统搜索无人机论文所使用的关键词。

acrobatic 杂技	bat 蝙蝠	flight 飞行	rotor 动盘
aerial 举高	bee 蜜蜂	fly 苍蝇	rotorcraft 旋翼机
aero 航空	bird 鸟儿	flying 飞行	soar 翱翔
aeroplane 飞机	blimp 飞艇	glide 牵引滑翔机	soaring 高扬
air 气	copter 直升机	glider 滑翔机	micro aerial vehicle 微型飞行器
aircraft 飞机	dragonfly 蜻蜓	gliding 滑翔	unmanned aerial vehicle 无人机
airplane 飞机	drone 嗡嗡声	hover 悬停	unmanned aircraft system 无人驾驶飞机系统
airship 飞船	flap 扑扇	hovering 婆娑	vertical takeoff and landing 垂直起降
balloon 球囊	flapping 拍打	kite 纸鸢	MAV, UAV, UAS, VTOL MAV、UAV、UAS、VTOL

of Robotics Research (IJRR), IAS Robotics and Autonomous Systems (RAS), IEEE Robotics and Automation Letters (RA-L), ACM Journal on Interactive, Mobile, Wearable and Ubiquitous Technologies (IMWUT), and ACM Transactions on Human-Robot Interaction (THRI); the seven conferences include IEEE International Conference on Intelligent Robots and Systems (IROS), IEEE International Conference on Robotics and Automation (ICRA), ACM/IEEE International Conference on Human-Robot Interaction (HRI), IEEE International Workshop on Robot and Human Communication (ROMAN), ACM Conference on Human Factors in Computing Systems (CHI), ACM International Conference on Ubiquitous Computing (UbiComp), and ACM International Conference on Human-Computer Interaction with Mobile Devices and Services (MobileHCI). We also manually reviewed the hard copies of the IROS and ICRA conferences' table of contents from 2001 to 2004, as we find that not all UAV papers in those years are listed on the website (IEEE Xplore).
机器人研究》（IJRR）、《机器人与自主系统》（RAS）、《IEEE 机器人与自动化通讯》（RA-L）、《ACM 交互、移动、可穿戴和泛在技术期刊》（IMWUT）和《ACM 人机交互期刊》（THRI）；七个会议包括：IEEE 智能机器人与系统国际会议 (IROS)、IEEE 机器人与自动化国际会议 (ICRA)、ACM/IEEE 人机交互国际会议 (HRI)、IEEE 机器人与人交流国际研讨会 (ROMAN)、ACM 计算系统中的人为因素会议 (CHI)、ACM 泛在计算国际会议 (UbiComp)，以及 ACM 移动设备与服务人机交互国际会议 (MobileHCI)。我们还手工查阅了 2001 年至 2004 年 IROS 和 ICRA 会议目录的打印文本，因为我们发现这些年的无人机论文并未全部列在网站（IEEE Xplore）上。

Then, we designed a list of keywords (Table 1) to search drone papers systematically from the titles and abstracts collected in the first step. Note that we searched for both the full name of each keyword (e.g., Unmanned Aerial Vehicle) and its abbreviation (i.e., UAV) with an automatic program script. The keywords include most of the words that describe a UAV. For example, the word "sUAV" (small UAV) could be detected by the keyword "UAV". Similarly, the word "quadcopter" or "quadrotor" could be detected by the keyword "copter" or "rotor". As long as one of the keywords is detected, the paper will pass this automated screening process.
然后，我们设计了一个关键词列表（表 1），从第一步收集的标题和摘要中系统地搜索无人机论文。请注意，我们使用自动程序脚本搜索了每个关键词的全称（如无人机）及其缩写（即 UAV）。这些关键词包括描述无人飞行器的大部分单词。例如，"sUAV"（小型无人飞行器）一词可由关键词 "UAV "检测出来。同样，"quadcopter"（四旋翼飞行器）或 "quadrotor"（四旋翼飞行器）也可以通过关键词 "copter"（直升机）或 "rotor"（旋翼）检测到。只要检测到其中一个关键词，论文就能通过自动筛选程序。
Finally, we performed a manual screening to reject some non-drone papers. We read the abstract, section titles, related works, and experiment results of all the papers from the second step. If a paper passes all the five criteria below, we consider it a drone paper for this survey.

最后，我们进行了人工筛选，剔除了一些非无人机论文。我们阅读了第二步中所有论文的摘要、章节标题、相关作品和实验结果。如果一篇论文通过了以下五项标准，我们就认为它是本次调查的无人机论文。

The paper must have more than two pages; we do not consider workshop and poster papers.
论文必须超过两页；我们不考虑研讨会和海报论文。
The paper must have at least one page of flightrelated results. These can be either simulation / experiment results, prototyping / fabrication results, or insights / discussion / lesson learned. One exception is a survey/review paper, which normally does not present experiment results. Papers with details or photos of the UAV hardware are a plus. Note that the experiment results do not necessarily need to be a successful flight, e.g., flapping wing UAVs normally have on-the-bench test results.
论文必须至少有一页是与飞行相关的结果。这些结果可以是模拟/实验结果、原型/制造结果，也可以是见解/讨论/经验教训。调查/综述类论文除外，这类论文通常不提供实验结果。附有无人机硬件细节或照片的论文更佳。请注意，实验结果不一定是成功的飞行，例如，拍翼无人机通常会有台上测试结果。
In topics related to computer vision, the images must be collected from a UAV's onboard camera rather than a manually moving camera.
在与计算机视觉相关的主题中，图像必须从无人机的机载相机而不是手动移动的相机中收集。
In topics related to computer vision, the images must be collected by the authors themselves. This is important, as authors who collected the dataset themselves often provide insights about their data collection and experiment results.
在与计算机视觉相关的主题中，图像必须由作者自己收集。这一点非常重要，因为亲自收集数据集的作者往往能提供有关其数据收集和实验结果的见解。
The paper which proposes a general method, e.g., path planning, must have related works and experiment results on drones. This is important, as some authors mention that their method can be applied to a UAV, but provide no experiment result to verify their statement.
提出通用方法（如路径规划）的论文必须有相关著作和无人机实验结果。这一点很重要，因为有些作者提到他们的方法可以应用于无人机，但却没有提供实验结果来验证他们的说法。

It is interesting to note that using the keyword "air" in the second step increases the number of false entries (since the keyword is used in many contexts) but helps to identify some rare drone-related papers that have only the keyword "air" in the title and abstract. By manually filtering the list in the third step, we successfully identified two of these drone papers 14 . Similarly, using the keyword "bee" can help to identify a rare drone paper [24. On the other hand, we chose not to use the keyword of "wing" because it causes many false entries like the case of "following", "knowing", etc.
值得注意的是，在第二步中使用关键词 "air "会增加错误条目的数量（因为该关键词被用于多种语境），但却有助于识别出一些罕见的无人机相关论文，这些论文的标题和摘要中只有关键词 "air"。通过在第三步中对列表进行人工筛选，我们成功识别出了其中两篇无人机论文14 。同样，使用关键词 "bee "也有助于识别罕见的无人机论文[24]。另一方面，我们选择不使用 "翅膀 "这一关键词，因为它会导致许多错误条目，如 "跟随"、"知道 "等。

4 Perceptual Map of UAVs
4 无人飞行器感知图

Having a framework that could cover all the related companion UAV works in both engineering and social
建立一个框架，涵盖无人机在工程和社会领域的所有相关配套工作

interaction topics is challenging, as these papers have different motivation, methodology, and results. Companion UAV works focusing on engineering usually emphasize on devising new hardware designs or new autonomous functions, while companion UAV works focusing on social interaction studies usually emphasize on participatory design and social experiments with users. To this end, we propose to categorize related works in this survey based on a perceptual map with the degree of autonomy (corresponding to engineering works) and the degree of sociability (corresponding to social interaction works) (Fig. 1)
由于这些论文的动机、方法和结果都不尽相同，因此研究交互主题具有挑战性。以工程学为重点的伴飞无人机作品通常强调设计新的硬件或新的自主功能，而以社会互动研究为重点的伴飞无人机作品通常强调参与式设计和与用户的社会实验。为此，我们建议根据自主程度（与工程作品相对应）和社会性程度（与社会互动作品相对应）的感知图，对本调查中的相关作品进行分类（图 1）。

4.1 The Four UAV Categories
4.1 四类无人机

The perceptual map of UAVs has four categories: remotecontrolled UAV, autonomous UAV, social UAV, and companion UAV. Traditionally, UAVs are controlled manually by human operators and have low degrees of autonomy and sociability. Gradually, along the vertical axis of degree of autonomy, researchers have been improving the autonomy aspects of UAVs, such as better reactive control with more sensors and better path planning algorithms. Essentially, autonomous UAVs are less dependent on human operators and are able to perform simple flight tasks autonomously.
无人机的感知图谱分为四类：遥控无人机、自主无人机、社交无人机和陪伴无人机。传统上，无人机由人类操作员手动控制，自主性和社交性较低。沿着自主程度的纵轴，研究人员逐渐改进了无人机的自主性，如利用更多传感器和更好的路径规划算法实现更好的反应控制。从根本上说，自主无人机对人类操作员的依赖程度较低，能够自主执行简单的飞行任务。

At the same time, along the horizontal axis of degree of sociability, researchers have been improving the social aspects of UAVs, such as designing UAV movements that are more comfortable for humans or building intuitive interfaces for us to understand UAVs' attention better. Most HRI researchers focus on the social aspects of UAVs and usually perform user studies using Wizard of

experiments. Different from autonomous UAVs, in which its main purpose is to achieve a task efficiently from an engineering point of view, social UAV aims to work with human harmonically, i.e., ease user acceptance and relieve user's cognitive burden. For example, a "social" fire-fighting drone might need to have a design that make nearby human understanding its purpose of fire-fighting during emergency 51].
与此同时，沿着社交程度的横轴，研究人员一直在改进无人机的社交方面，例如设计让人类更舒适的无人机动作，或为我们更好地理解无人机的注意力建立直观的界面。大多数人机交互研究人员都关注无人机的社交方面，通常使用

实验向导进行用户研究。从工程学的角度来看，自主无人机的主要目的是高效地完成任务，而社交无人机则不同，它的目的是与人类和谐共处，即减轻用户的认知负担，让用户更容易接受。例如，"社会化 "消防无人机的设计可能需要让附近的人类理解它在紧急情况下灭火的目的 51]。

We first coined the phrase "companion UAV" and consider a companion UAV as one that possesses high degrees of both autonomy and sociability 62 . In addition to the autonomy skills, such as motion planning and obstacle avoidance, companion UAVs must also feature sociability skills such as making users feel safe and understand their intention. It is worth noting that the term "companion UAV" used in a prior work 41]
我们首次创造了 "陪伴型无人机 "这一短语，并将陪伴型无人机视为同时具备高度自主性和社交性的无人机 62 。除了运动规划和避障等自主技能外，陪伴型无人机还必须具备社交技能，例如让用户感到安全并理解他们的意图。值得注意的是，之前的一项研究使用了 "陪伴型无人机 "这一术语 41] 。

has a different meaning, where the UAV was designed to support a ground robot but not to interact with a person.
其含义不同，无人飞行器的设计目的是为地面机器人提供支持，而不是与人互动。

To consolidate the idea of perceptual map, we can use a package-delivery drone as an example. In the autonomous UAV sense, the package-delivery drone focuses on accomplishing the delivery task from the engineering perspective. Note that this alone is a challenging task as one needs to figure out how to perform the flight efficiently, how to detect the landing location, how to avoid obstacles during the flight, etc. On the other hand, in the sociable UAV sense, the package-delivery drone should also acknowledge people for successful interactions with people, i.e., signaling to the person that it has seen them at a certain distance and a range of time. This design aspect has also been raised and investigated recently by Jensen et al. 46 .
为了巩固 "感知地图 "的概念，我们可以以无人机包裹递送为例。在自主无人机的意义上，包裹递送无人机侧重于从工程角度完成递送任务。需要注意的是，这本身就是一项具有挑战性的任务，因为我们需要弄清楚如何高效地执行飞行、如何检测着陆位置、如何在飞行过程中避开障碍物等。另一方面，从交际无人机的意义上讲，包裹递送无人机还应在与人成功互动时对人表示认可，即向对方发出信号，表明它在一定距离和时间范围内看到了对方。Jensen 等人最近也提出并研究了这一设计方面的问题46 。

4.2 Link with Previous Study
4.2 与以往研究的联系

It should be noted that "social UAV" has been mentioned in the past literature frequently. One representative example is an survey paper of social UAV 6]. In this study, Baytaş et al. define "social drone" as autonomous flyers operate near to human users. Literally, this definition is similar to the definition of "companion UAV" here, but upon careful investigation, their true meaning of "social drone" is closer to the meaning of "social UAV" mentioned in the perceptual map here. Most papers considered by Baytaş et al. were not truly autonomous, i.e., a drone with pre-programmed behavior and motions is considered as autonomous by them. In our opinion, their categorization is not precise enough, e.g., while the teleoperated drone 49 is categorized as a "social drone" by them, we consider that teleoperated drone a remotely-controlled drone in our context here as the drone is neither autonomous nor social. They also included poster and short papers in their review, but it is unclear how they categorize some short and poster papers that lack of implementation details. In contrast, our work here cover more papers in a more coordinated and systematic way.
值得注意的是，"社会无人机 "在过去的文献中经常被提及。一个有代表性的例子是一篇关于社交无人机的调查论文 6]。在这项研究中，Baytaş 等人将 "社交无人机 "定义为在人类用户附近运行的自主飞行器。从字面上看，这一定义与本文对 "陪伴型无人机 "的定义相似，但仔细研究后发现，他们对 "社交型无人机 "的真正理解更接近本文感知图谱中提到的 "社交型无人机 "的含义。巴伊塔斯等人所考虑的大多数论文都不是真正意义上的自主论文，也就是说，他们认为具有预编程行为和动作的无人机是自主的。我们认为，他们的分类不够准确，例如，他们将远程操作的无人机 49 归类为 "社交无人机"，但我们认为，在我们这里，远程操作的无人机是遥控无人机，因为该无人机既不是自主的，也不是社交的。他们的综述还包括海报论文和短篇论文，但不清楚他们是如何对一些缺乏实施细节的短篇论文和海报论文进行分类的。相比之下，我们的工作以更加协调和系统的方式涵盖了更多论文。

4.3 Research Efforts Towards Companion UAVs
4.3 针对配套无人机的研究工作

Designing companion UAVs is challenging as it involves both technical/engineering issues and social/emotional design questions. We believe that because of this reason, most UAV works identified in this survey focus on a single issue, either an engineering or social issue, rather than having an ambitious goal to tackle both issues in a paper. As shown in Fig. 2, there are two main efforts
设计陪伴型无人机具有挑战性，因为它既涉及技术/工程问题，也涉及社会/情感设计问题。我们认为，正因为这个原因，本次调查中发现的大多数无人机作品都只关注一个问题，即工程问题或社会问题，而不是雄心勃勃地在一篇论文中同时解决这两个问题。如图 2 所示，主要有两方面的工作

Fig. 2 Perceptual map for UAVs based on the degree of autonomy and sociability, with the major topics found in the literature.
图 2 基于自主性和社会性程度的无人飞行器感知图，以及文献中发现的主要主题。

for realizing companion UAVs in the literature, where the first one moves from the remote-controlled UAV to the autonomous UAV direction (blue arrow), and the second one moves from the remote-controlled UAV to the social UAV direction (red arrow).
其中，第一个箭头从遥控无人机向自主无人机方向移动（蓝色箭头），第二个箭头从遥控无人机向社会无人机方向移动（红色箭头）。

The blue arrow in Fig. 2 signifies efforts of robotic developers in realizing companion UAVs. In the literature, these efforts have a distinctive feature where the authors include engineering details, be it about the UAV hardware, control algorithms, or visual tracking methods. From the identified companion UAV works in Section 3 , the topics of human-following UAVs and user interface clearly emerge in this area. In Section 5, we will discuss these sub-topics in more details.
图 2 中的蓝色箭头表示机器人开发人员在实现伴飞无人机方面所做的努力。在这些文献中，作者们将无人机硬件、控制算法或视觉跟踪方法等工程细节纳入其中。从第 3 节中确定的陪伴式无人机作品来看，人类跟随无人机和用户界面显然是这一领域的主题。在第 5 节中，我们将更详细地讨论这些子课题。

The red arrow in Fig. 2 signifies efforts of HDI researchers in realizing companion UAVs. In the literature, these efforts have a distinctive feature where the authors performed Wizard-of-Oz experiments or carried out online surveys by using HDI videos. From the identified companion UAV works in Section 3, the topics of social perception of UAVs, emotion and intention expression of UAVs, gesture and physical interaction with UAVs clearly emerge in this area. In Section 6e will discuss these sub-topics in more details.
图 2 中的红色箭头表示人类发展倡议研究人员在实现配套无人机方面所做的努力。在这些文献中，作者们利用 HDI 视频进行了 Wizard-of-Oz 实验或在线调查。从第 3 节中确定的陪伴式无人机作品来看，该领域的主题明显包括无人机的社会感知、无人机的情感和意图表达、与无人机的手势和肢体交互。第 6e 节将详细讨论这些子课题。

5 From Remote-Controlled to Autonomous UAVs
5 从遥控无人机到自主无人机

Developing companion UAVs that are autonomous and sociable is not a straightforward task. Most companion UAV works focus on one topic for realizing a companion UAV. In this section, we aim to summarize engineering efforts for realizing a companion UAV, with focus on the human-following and control interface topics.
开发可自主飞行且具有社交能力的伴飞无人机并不是一项简单的任务。大多数陪伴型无人机的研究都集中在实现陪伴型无人机的一个主题上。在本节中，我们将总结为实现陪伴型无人机所做的工程努力，重点关注人类跟随和控制界面主题。

5.1 Human Following UAVs 5.1 人类跟踪无人飞行器

Pestana et al. used a UAV's onboard camera and object tracking algorithm to realize a human following application 88. Higuchi et al. also performed human following with the UAV's onboard camera by using a color-based particle filter 40. On the other hand, Papachristos et al. demonstrated a human tracking application with the UAV's onboard stereo camera 83. All the proposed prototypes focused on the functional designs of the system and did not carry out social interaction experiments. Moreover, these systems have special requirements such as manual initialization of the user location 88, the necessity for the user to wear a shirt of a specific color [40], or the necessity for the user to move (so that the image tracker starts working) 83.
Pestana 等人利用无人机的机载摄像头和物体跟踪算法实现了人体跟踪应用 88。Higuchi 等人也利用无人飞行器的机载摄像头，通过使用基于颜色的粒子滤波器 40 实现了人类跟踪。另一方面，Papachristos 等人利用无人机的机载立体相机 83 演示了人体跟踪应用。所有提出的原型都侧重于系统的功能设计，并没有进行社交互动实验。此外，这些系统都有一些特殊要求，如手动初始化用户位置 88、用户必须穿特定颜色的衬衫 [40]，或用户必须移动（以便图像跟踪器开始工作）83。

By integrating a visual SLAM technique and a visionbased human tracking algorithm, Lim & Sinha presented a UAV that can map the human walking path in real time 65]. On the other hand, Nasser et al. proposed a UAV that can perform human following and gesture recognition with an onboard Xtion depth camera 76 . More recently, Yao et al. integrated a face detector and a feature tracker in order to achieve robust human tracking with a miniature robotic blimp 117. Note that these systems are not able to track and follow the user robustly in every situation, e.g., when the user is occluded by other objects/people. In order to tackle the occlusion problem, Hepp et al. presented a human-following system based on ultra-wideband (UWB) radio and released their implementation as open-source software 39.
通过整合视觉 SLAM 技术和基于视觉的人体跟踪算法，Lim 和 Sinha 提出了一种可实时绘制人体行走路径的无人机 65]。另一方面，Nasser 等人提出了一种无人机，可通过机载 Xtion 深度相机进行人体跟踪和手势识别 76。最近，Yao 等人集成了人脸检测器和特征跟踪器，利用微型机器人飞艇实现了稳健的人体跟踪 117。需要注意的是，这些系统并不能在任何情况下都能稳健地跟踪用户，例如，当用户被其他物体/人遮挡时。为了解决遮挡问题，Hepp 等人提出了一种基于超宽带（UWB）无线电的人体跟踪系统，并将其实现作为开源软件发布 39。

A few works on human-following UAV focus on filming. Huang et al. combined a stereo camera-based human following method with a dynamic planning strategy to film a human action in a more expressive manner 43. Zhou et al. designed a flying drone that could keep tracking a human motion using a normal color camera 121. Bentz et al. presented an assistive aerial robot that could observe regions most interesting to the human and broadcast these views to the humans augmented reality display 77 . This resulted in reduced head motions of the human as well as improved reaction time.
少数几项关于人类跟随无人机的研究侧重于拍摄。Huang 等人将基于立体相机的人体跟随方法与动态规划策略相结合，以更富表现力的方式拍摄人体动作 43。Zhou 等人设计了一种飞行无人机，可使用普通彩色摄像机持续跟踪人的动作 121。Bentz 等人提出了一种辅助飞行机器人，它可以观察人类最感兴趣的区域，并将这些景象播放到人类的增强现实显示屏 77 上。这减少了人类的头部运动，并改善了反应时间。

5.2 User Interfaces for UAVs
5.2 无人飞行器的用户界面

In 2013, Monajjemi et al. presented a method to command a team of UAVs by using face and hand gestures 71. Later, Monajjemi et al. extended their work by commanding a team of two UAVs using not only face engagement and hand gestures but also voice and touch interfaces 70. Similar to Monajjemi's works on multi-modal interaction above, MohaimenianPour & Vaughan 69] and Nagi et al. 75] realized UAV control with hands and faces by relying on visual object detectors and simple preset rules
2013 年，Monajjemi 等人提出了一种通过面部和手势指挥一队无人机的方法 71。后来，Monajjemi 等人扩展了他们的工作，不仅使用面部参与和手势，还使用语音和触摸界面来指挥由两架无人机组成的团队 70。与 Monajjemi 的上述多模态交互工作类似，MohaimenianPour & Vaughan 69] 和 Nagi 等人 75] 依靠视觉对象检测器和简单的预设规则，实现了用手和脸控制无人飞行器

Unlike Monajjemi's works on multi-modal interaction above, Sun et al. focused on piloting a drone with gesture recognition by combining a visual tracker with a skin pixel detector for robust performance 102. Similarly, Lichtenstern et al. demonstrated a system where a user can control multiple UAVs using hand gestures 61.
与 Monajjemi 的上述多模态交互作品不同，Sun 等人的研究重点是通过手势识别来驾驶无人机，方法是将视觉跟踪器与皮肤像素检测器结合起来，以获得强大的性能 102。同样，Lichtenstern 等人展示了一个用户可以通过手势 61 控制多架无人机的系统。

Constante et al. aimed to improve the hand gesture interface of UAVs by proposing a new algorithm transfer learning algorithm that can exploit both online generic and user-specific hand gestures data 23. Burke

Lasenby presented a very fast and simple classification method to control a UAV with pantomimic gestures, in which the main idea to use a gesture that is similar to the desired action of UAV as a gesture command 13 .
Constante等人旨在通过提出一种新的算法转移学习算法来改进无人机的手势界面，该算法可以利用在线通用手势数据和用户特定手势数据 23。Burke

Lasenby 提出了一种非常快速和简单的分类方法，利用拟态手势控制无人机，其主要思想是使用与无人机所需动作相似的手势作为手势指令 13。

More recently, Bruce et al. proposed the use of facial expression for 3D trajectory control of UAVs 12. Previously, we have also demonstrated a drone that could react to the user's facial expression [64]. In contrast to facial expression, Huang et al. directed a UAV in a known environment via natural language commands 42 .
最近，布鲁斯等人提出将面部表情用于无人机的三维轨迹控制 12。在此之前，我们也展示过一种能对用户面部表情做出反应的无人机 [64]。与面部表情不同，Huang 等人通过自然语言指令在已知环境中指挥无人机 42 。

This section summarizes efforts in sociability studies for realizing companion UAVs. This section also offers different perspective from the recent survey work of social drones 6]. We first discuss the social perception of UAVs, followed by topics on emotion and intention expression of UAVs through motions, lights, or displays. Then, we briefly describe related works in gesture interaction and physical interaction with UAVs.
本节总结了为实现陪伴型无人机的社会性研究而做出的努力。本节还提供了与近期社交无人机调查工作不同的视角 6]。我们首先讨论无人机的社交感知，然后讨论无人机通过动作、灯光或显示屏表达情感和意图的话题。然后，我们简要介绍了与无人机进行手势交互和物理交互的相关工作。

Designing companion UAVs which invite social interaction is important. Wojciechowska et al. investigated the best way for a flying robot to approach a person 115. Yeh et al. found that a drone with circular body shape, face, and voice could reduce the proximate distance between a social drone and the user 118. In addition, there are also studies on user perception on UAV, focusing on assistance during emergency situations 52, privacy and security issues 19, and autonomous behaviors 80 .
设计能够吸引社交互动的陪伴型无人机非常重要。Wojciechowska 等人研究了飞行机器人接近人的最佳方式 115。Yeh 等人发现，带有圆形体形、面部和声音的无人机可以缩短社交无人机与用户之间的近距离 118。此外，还有关于用户对无人机感知的研究，主要集中在紧急情况下的协助 52、隐私和安全问题 19 以及自主行为 80 等方面。

Different from the social interaction works mentioned above, Abtahi et al. explored the touch interaction in HDI and participants preferred interacting with a safeto-touch drone in the studies [2]. In particular, users feel safer and were less mentally demanding when interact with the safe-to-touch drone.
与上述社交互动作品不同，Abtahi 等人探索了人类发展倡议中的触摸互动，参与者在研究中更喜欢与安全触摸无人机互动[2]。特别是，在与安全触摸无人机互动时，用户感觉更安全，对精神的要求也更低。

6.2 Emotion and Intention Expression of UAVs
6.2 无人飞行器的情感和意图表达

Dancers use various kinds of motion to express their emotions. Sharma et al. used Laban motion analysis (a common method used by artists to express emotions) for UAVs to express their affective states 96. Aiming to deliver an opera performance, Eriksson et al. also described their method of designing expressive motions interactively with a choreographer for drones 29 .
舞蹈家使用各种运动来表达他们的情感。Sharma 等人在无人机上使用拉班动作分析法（艺术家表达情感的常用方法）来表达他们的情感状态 96。为了进行歌剧表演，Eriksson 等人还介绍了他们与编舞家互动设计无人机表情动作的方法 29。

Similarly, Cauchard et al. presented a model for UAVs to express emotions via movements [18, believing that encoding these emotions into movements could help users to comprehend the UAV's internal states.
同样，Cauchard 等人提出了一个无人机通过动作表达情感的模型[18]，认为将这些情感编码成动作可以帮助用户理解无人机的内部状态。

In contrast to emotion expression, Szafir et al. used the UAV's motion to express the robot's intention 103 . Walker et al. expand this work by visualizing robot motion intent using an augmented reality technique 113 . Colley et al. also investigated drone motion as direct guidance for pedestrians rather than equipping drones with a display or indicators 21.
与情感表达不同，Szafir 等人利用无人机的运动来表达机器人的意图 103 。沃克等人利用增强现实技术将机器人的运动意图可视化，从而扩展了这项工作 113 。Colley 等人也研究了无人机运动对行人的直接引导，而不是为无人机配备显示屏或指示器 21。

Duncan et al. have similar idea and presented an initial study for UAVs to communicate their internal states to bystanders via flying patterns [26]. In their seminal work, Firestone et al. performed a participatory design with users for UAVs to communicate internal states effectively via flying patterns 31.
Duncan 等人也有类似的想法，并提出了一项关于无人机通过飞行模式向旁观者传达其内部状态的初步研究[26]。在他们的开创性工作中，Firestone 等人与用户一起进行了参与式设计，让无人机通过飞行模式有效地传达内部状态[31]。

LED light has also been used for UAVs to express their emotion and intent. Arroyo et al. described a social UAV that performs four different expressions with head movement and two color LED eyes 5. Szafir et al. also used a ring of sixty-four color LEDs as a reliable cue for the UAV to convey intention to the user 104.
LED 灯还被用于无人机表达情感和意图。Arroyo 等人描述了一种社交无人机，该无人机通过头部运动和两只彩色 LED 眼睛 5 表现出四种不同的表情。Szafir 等人也使用了一个由六十四颗彩色 LED 组成的环作为无人机向用户传达意图的可靠提示104。

Instead of using LED light, some works rely on displays or projectors to convey information to users, including a small drone with an OLED display for telepresence function HDI 35, a flying display system for crowd control during emergency situations [95], a flying projector-screen system with two UAVs 81, and flying UAVs with onboard projectors for social group interactions [94], interactive map application [11], navigation guidance 56], and gesture interaction (17.
一些作品没有使用 LED 灯，而是依靠显示器或投影仪向用户传递信息，包括用于远程呈现功能的带 OLED 显示屏的小型无人机 HDI 35、用于紧急情况下人群控制的飞行显示系统[95]、带两个无人机的飞行投影仪屏幕系统 81，以及用于社会群体互动的带机载投影仪的飞行无人机[94]、交互式地图应用[11]、导航引导 56]和手势互动（17.

6.3 Gesture Interaction with UAVs
6.3 与无人飞行器的手势交互

Inspired by human interaction with birds, Ng & Sharlin studied the effectiveness of a few hand gestures in commanding a UAV [77]. Participants were very engaged when having gesture interaction with the UAV and spoke to the UAV like a pet. Cauchard et al. also performed similar Wizard-of-Oz experiments and most participants interacted with the UAV as if it were a pet 16. E et al. later expand this experiment in different culture setting and found similar results 28].
受人类与鸟类互动的启发，Ng 和 Sharlin 研究了一些手势在指挥无人机方面的有效性[77]。参与者在与无人飞行器进行手势互动时非常投入，并像对待宠物一样与无人飞行器交谈。Cauchard 等人也进行了类似的 "Wizard-of-Oz "实验，大多数参与者都像对待宠物一样与无人机进行互动[16]。E 等人后来在不同的文化背景下扩展了这一实验，发现了类似的结果 28]。

Aiming to increase the naturalness of HDI, Peshkova et al. surveyed gesture interaction techniques that have been applied for UAV control based on three mental models: the imitative class (controls the UAV motions with the user's body movements), the instrumented class (controls the UAV motions with a physical controller or an imaginary object), and the intelligent class (interacts with a UAV as if the UAV is an intelligent agent) 87.
为了提高人类发展指数的自然度，Peshkova 等人研究了基于三种心理模型的无人机控制手势交互技术：模仿类（用用户的肢体动作控制无人机运动）、工具类（用物理控制器或假想物体控制无人机运动）和智能类（将无人机作为智能代理与之交互）87。

On the other hand, Pfeil et al. studied the effectiveness of different interaction techniques of the upper body in UAV control 89 (including all the three interaction classes mentioned by Peshkova et al. 87]). They found that the proxy technique, in which the user moves the UAV as he/she is grasping the UAV in his/her hand, is the best out of the five developed interaction techniques.
另一方面，Pfeil 等人研究了无人机控制中上半身不同交互技术的有效性 89（包括 Peshkova 等人提到的所有三种交互类别 87]）。他们发现，在已开发的五种交互技术中，代理技术（即用户在手握无人机时移动无人机）效果最好。

6.4 Physical Interaction with UAVs
6.4 与无人飞行器的物理交互

Physical interaction is rare in the UAV literature compared to gesture interaction. Knierim et al. used physical interaction with a flying robot as a novel input (touch/drag the flying robot) and output (the flying robot generates forces feedback) modalities for a user interface 55. Abtahi et al. proposed a haptic interaction

system, where an actual UAV is used to enhance user's physical perception in virtual reality environment (1]. Soto et al. explored the idea of using a leashed UAV as a navigator to guide visually impaired people 99 .
与手势交互相比，物理交互在无人机文献中并不多见。Knierim 等人将与飞行机器人的物理交互作为一种新颖的输入（触摸/拖拽飞行机器人）和输出（飞行机器人产生力反馈）模式，用于用户界面 55。Abtahi 等人提出了一种触觉交互

系统，利用实际的无人飞行器来增强用户在虚拟现实环境中的物理感知（1]。Soto 等人探讨了使用拴有绳索的无人机作为导航仪为视障人士提供导航的想法 99 。

7 Observation and Lessons Learned
7 意见和经验教训

Throughout the review process and personal experience, we noticed several patterns in the literature and learned a few lessons in designing companion UAVs. We discuss these observation (including ideas exploration) in this section, including: (i) UAV form, (ii) appearance design, (iii) integrated human-accompanying model, (iv) integrated human-sensing interface, (v) safety concerns, (vi) noise issue and sound design, and tactile interaction. Note that several aspects mentioned in this section
通过查阅文献和亲身经历，我们注意到文献中的一些模式，并从中汲取了一些设计陪伴型无人机的经验教训。我们将在本节中讨论这些观察结果（包括想法探索），包括：(i) 无人机外形，(ii) 外观设计，(iii) 综合人类陪伴模型，(iv) 综合人类传感界面，(v) 安全问题，(vi) 噪音问题和声音设计，以及触觉交互。请注意，本节中提到的几个方面

\footnotetext{ \脚注文本{

Physical HDI with a virtual reality display in their context.

实体人类发展指数与虚拟现实显示屏的结合。

could be potentially improved by drawing inspiration from the human-computer interaction or human-robot interaction literature. In next section, we will present a more concise guidelines and recommendations towards realizing companion UAVs.
通过从人机交互或人机交互文献中汲取灵感，我们有可能改进 "陪伴式无人机 "的设计。在下一节中，我们将为实现陪伴型无人机提出更简明的指导原则和建议。

7.1 UAV Form 7.1 无人飞行器形式

Almost all papers considered in this work use multirotor UAV as a platform to demonstrate UAV flight or to carry out social experiments. From our long experience working with drones, we agree that multirotors are more convenient for experiments (in term of availability) and presentation (in term of flight quality) but their noise level is too annoying for companion UAVs.
几乎所有论文都将多旋翼无人机作为展示无人机飞行或进行社会实验的平台。根据我们长期使用无人机的经验，我们一致认为多旋翼无人机在实验（可用性方面）和演示（飞行质量方面）方面更加方便，但其噪音水平对于伴飞无人机来说过于恼人。

We argue that a blimp or balloon type UAV is more suitable as a form for companion UAVs. We list up two technical blimp design papers that could be an alternative form of companion UAVs. First, Song et al. used a novel idea to hide the propulsion unit in the center of a blimp and designed a blimp that is safe to touch and interact with 98. Second, Yamada et al. designed a blimp with micro-blower with no rotating blade (hence safer and quieter) 116. It is also worth mentioned that Drew et al. designed a small flying robot using electrohydrodynamic thrust with no moving part but it is tiny and cannot handle large payload. 25
我们认为，飞艇或气球型无人机更适合作为伴飞无人机的形式。我们列举了两篇技术性飞艇设计论文，它们可以作为伴飞无人机的另一种形式。首先，Song 等人采用了一种新颖的想法，将推进装置隐藏在飞艇中心，并设计了一种可安全触摸和与 98 互动的飞艇。其次，Yamada 等人设计的飞艇带有微型鼓风机，没有旋转叶片（因此更安全、更安静） 116。值得一提的还有 Drew 等人设计的一种小型飞行机器人，它使用电流体动力推力，没有移动部件，但体积很小，无法处理大的有效载荷。25

7.2 Appearance Design 7.2 外观设计

Appearance design of drones is important as the design affect users' perception 114. A few HDI studies mentioned in Section 6.1 investigated the user perception on drones. For example, it is found that a round shape flying robot has higher social acceptance 118 and emergency response drones with a prominent appearance that is easily distinguishable from recreational drones can gain user trust 52 . HDI researchers have mentioned about the importance of color in drone design 19. At the time of writing, no study has investigated the color design in companion UAVs and the most related study we can find is using color of balloons to visualize their surrounding air quality 58 .
无人机的外观设计非常重要，因为设计会影响用户的感知。第 6.1 节提到的一些人类发展倡议研究调查了用户对无人机的看法。例如，研究发现，圆形飞行机器人具有更高的社会接受度 118，应急无人机的外观突出，容易与休闲无人机区分开来，可以获得用户的信任 52。人类发展倡议的研究人员提到了颜色在无人机设计中的重要性 19。在撰写本文时，还没有研究调查过配套无人机的色彩设计，我们能找到的最相关的研究是利用气球的颜色来直观显示其周围的空气质量 58 。

7.3 Human Accompanying Model
7.3 人类陪伴模式

A few papers demonstrated human-following capability of companion UAVs (Section 5.1) and Wojciechowska et al. investigated the best way a flying robot should approach a person 115. We noticed that there is a lack of a general model to unify various human accompanying
有几篇论文展示了伴飞无人机的人类跟随能力（第 5.1 节），Wojciechowska 等人研究了飞行机器人接近人类的最佳方式115。我们注意到，目前还缺乏一个通用模型来统一各种人类陪伴模式。
behaviors of companion UAVs, including approaching, following, side-by-side walking, leading or guiding, and flying above the user (to help observing things far away). This observation is also applicable to ground robots. With a unified human accompanying model, companion UAVs are expected to be able to truly accompany and interact with a human more naturally. For more details, we have summarized related works of various human accompanying modes of both flying robots and mobile robots in our previous work 62 .
同伴无人机的行为，包括接近、跟随、并排行走、引领或指导，以及在用户上方飞行（帮助观察远处的事物）。这种观察也适用于地面机器人。有了统一的人类陪伴模型，陪伴型无人机有望真正陪伴人类，并与人类更自然地互动。关于更多细节，我们在以前的工作中总结了飞行机器人和移动机器人的各种人类陪伴模式的相关工作62。

7.4 Human Sensing Interface
7.4 人体感应界面

Human can achieve natural interaction with each other using face, gesture, touch, and voice modalities simultaneously. A few papers demonstrated HDI with multiple modalities but most papers focus on a single modality. It is not straightforward to realize a companion UAV with all modalities since researches/engineers often focus on methods with a single modality. Recently, great effort in the deep learning topic has led to a more integrated human sensing interface, such as the OpenPose library integrates visual tracking of human body, face, and hand simultaneously 15 . We should leverage these powerful tools more in order to extent companion UAV research. As a good example, Huang et al. utilized the OpenPose library for human-following and realize a high-level autonomous filming function with their UAV 43. A standard human sensing interface (could be updated regularly) is crucial in facilitating a HDI study, not only it can accelerate HDI progress, but also make comparison study more effective.
人类可以同时使用面部、手势、触摸和语音模式实现自然交互。有几篇论文展示了具有多种模式的人机交互，但大多数论文只关注单一模式。由于研究人员/工程师通常只关注单一模式的方法，因此要实现具有所有模式的无人机伴侣并不简单。最近，深度学习领域的巨大努力带来了更加集成的人体传感界面，例如 OpenPose 库同时集成了人体、面部和手部的视觉跟踪15 。我们应更多地利用这些强大的工具，以拓展无人机伴飞研究。作为一个很好的例子，Huang 等人利用 OpenPose 库进行人体跟踪，并通过他们的无人机实现了高级自主拍摄功能 43。一个标准的人类感应界面（可定期更新）对于促进人类发展指数研究至关重要，它不仅能加快人类发展指数的进展，还能使对比研究更加有效。

7.5 Safety Concerns 7.5 安全问题

HDI safety is an important design aspects of companion UAV. Most UAVs described in this work have sharp rotating propellers that could injure nearby human, especially human eyes - such accidents have been observed in a medical case report 73 and a formal news 107. Existing solutions

include using ring-shape protectors around the propellers 4,85 , having net cases that fully cover the propellers

, or designing a cage to cover the entire drone

but these modifications worsen the flight efficiency and shorten the flight time (due to the increased payloads).
人类发展指数安全是伴飞无人机的一个重要设计方面。这项工作中描述的大多数无人机都有锋利的旋转螺旋桨，可能会伤到附近的人，尤其是人的眼睛--在一份医学病例报告73和一则正式新闻107中都观察到了此类事故。现有的解决方案

包括在螺旋桨周围使用环形保护装置 4,85，使用完全覆盖螺旋桨的网壳

，或设计一个笼子覆盖整个无人机

，但这些改装会降低飞行效率并缩短飞行时间（由于增加了有效载荷）。

Recently, Lee et al. claimed that a Coanda UAV has several advantages over standard UAVs, such as crash resistance and flight safety, thanks to its unique mechanical design 60. A UAV with flexible structures
最近，Lee 等人声称，由于其独特的机械设计，Coanda 无人机与标准无人机相比具有若干优势，如抗坠毁性和飞行安全性。具有柔性结构的无人飞行器

\footnotetext{ \脚注文本{

Including commercial UAV examples.

包括商用无人机实例。

could be less harmful when it unavoidably hit a user (as the UAV structure will absorb the crash impact). Based on this idea, UAVs with soft body frame 92 and flexible propellers 45 have been proposed.
当无人驾驶飞行器不可避免地撞击用户时，其危害性可能较小（因为无人驾驶飞行器的结构将吸收撞击力）。基于这一想法，有人提出了配备软体框架 92 和柔性螺旋桨 45 的无人飞行器。

In addition to the physical safety, making users feel safer (less cognitive burden) is also important. From our experience, most users are afraid that the UAV is going to crash when it starts to move (because unlike a common ground vehicle, a conventional UAV has to tilt or roll in order to move). We then designed a special type of drone - a holonomic UAV, where it can move horizontally without tilting, and the users expressed that the holonomic flight makes them feel safer. While several holonomic UAVs exists

there is no formal HDI study of holonomic UAV so far to the best of our knowledge.
除了物理安全，让用户感觉更安全（减少认知负担）也很重要。根据我们的经验，大多数用户担心无人机开始移动时会坠毁（因为与普通地面车辆不同，传统无人机必须倾斜或翻滚才能移动）。于是，我们设计了一种特殊类型的无人机--全息无人机，它可以水平移动而无需倾斜，用户表示全息飞行让他们感觉更安全。据我们所知，虽然目前已有几种符合人体工程学的无人机

，但迄今为止还没有对符合人体工程学的无人机进行正式的人类发展指数研究。

7.6 Noise Issue and Sound Design
7.6 噪音问题和音响设计

In addition to the safety concerns, noise issue is also important for HDI. Most UAVs produce unwanted noise with their high speed and high power rotating propellers. In our test, the noise of a commercial UAV 86 was measured to be as high as 82 dB one meter away, and is very close to the hazardous level of 85 dB as legislated by most countries 48]. Studies also suggested that noise has a strong association with health issues 88 and increased risk of accidents 68. These findings suggest that noise issue should be seriously considered in HDI.
除了安全问题，噪音问题对人类发展倡议也很重要。大多数无人机的高速和大功率旋转螺旋桨会产生不必要的噪音。在我们的测试中，商用无人机 86 的噪音在一米之外被测得高达 82 分贝，非常接近大多数国家立法规定的 85 分贝的危险水平 48]。研究还表明，噪声与健康问题88 和事故风险增加68 密切相关。这些研究结果表明，人类发展倡议应认真考虑噪音问题。

Sound design 66] is also a related topic for HDI. For example, car manufacturers harmonically tune the engines' noise so that their cars can sound more comfortable to the users 53. Norman discussed about the emotional association of sounds and everyday products in his book [79]. For example, an expensive melodious kettle (when the water is boiling) and a Segway self-balancing scooter (when the motor is rotating) sound exactly two musical octaves apart, making the unwanted noises sound like music. In the robotics context, Moore et al. utilized the motor noise to facilitate user interaction 72 and Song & Yamada proposed to express emotions through sound, color, and vibrations 97.
声音设计 66] 也是人类发展指数的一个相关主题。例如，汽车制造商会对发动机的噪音进行谐调，使汽车听起来让用户更舒服 53。诺曼在他的著作[79]中讨论了声音与日常产品的情感联系。例如，昂贵而悠扬的水壶（水沸腾时）和赛格威（Segway）自平衡滑板车（马达转动时）的声音正好相差两个八度音程，这使得不需要的声音听起来像音乐。在机器人领域，Moore 等人利用马达噪音促进用户互动 72，Song 和 Yamada 提议通过声音、颜色和振动表达情感 97。

The use of non-vocal, non-verbal, or non-linguistic utterance (NLU) for HDI is also a potential way to enhance a UAV's characteristics and expressiveness. It should be noted that NLU might be more useful than speech during a HRI flight as the propeller noise make speech recognition difficult. In movies, robots such as R2-D2 also use non-verbal utterance to enhance its communication with characters. For more details on the
在人类发展倡议中使用非声音、非语言或非语言语句（NLU）也是增强无人机特性和表现力的一种潜在方法。值得注意的是，在人类发展指数飞行过程中，由于螺旋桨噪音导致语音识别困难，因此非语言单位可能比语音更有用。在电影中，R2-D2 等机器人也使用非言语语句来加强与角色的交流。有关

study of NLU on HRI, readers are recommended to read the PhD thesis of Read 91.
建议读者阅读 Read 91 的博士论文。

7.7 Tactile Interaction
7.7 触觉交互

Peña & Tanaka proposed the use of a robot's body temperature to express emotional state [82]. Park & Lee also studied the effect of temperature with a companion dinosaur robot and found that skin temperature significantly affects users' perception [84]. It would be interesting and useful to explore this new thermal application area. For example, companion UAVs with warm/hot body frame could signify a status of hardworking/exhaustion.
Peña 和 Tanaka 提出利用机器人的体温来表达情绪状态 [82]。Park & Lee 也研究了温度对恐龙机器人伴侣的影响，发现皮肤温度会显著影响用户的感知[84]。探索这一新的热应用领域将是有趣而有益的。例如，伴行无人机的体温/体温框架可以表示一种辛劳/疲惫的状态。

Equipping UAVs with touch sensing capability allows richer HDI. For instance, a UAV could perceive a person's love if it could sense the person's gentle stroke. Prior studies and results in the social robotics field [119] could be integrated into a UAV to realize a more personalized companion UAV. In addition, from the engineering perspective, a UAV with touch sensors on its propeller guard could also sense a nearby human instantly and enhance HDI safety.
为无人飞行器配备触摸感应功能可实现更丰富的人类发展指数。例如，如果无人机能感知到一个人轻轻的抚摸，那么它就能感知到这个人的爱。社交机器人领域的前期研究和成果[119]可以整合到无人机中，从而实现更加个性化的陪伴型无人机。此外，从工程学的角度来看，在螺旋桨护罩上安装触摸传感器的无人机还可以即时感知附近的人类，提高人类发展倡议的安全性。

8 Guidelines and Recommendations
8 指导方针和建议

After mentioning several observation in the companion UAV literature (and our experience in designing companion UAVs), this section presents several design and research recommendations for companion UAVs. Note that almost all the topics discussed below are applicable to both the engineering development and social interaction experiments.
本节在介绍了同伴无人飞行器文献中的一些观察结果（以及我们在设计同伴无人飞行器方面的经验）之后，提出了一些关于同伴无人飞行器的设计和研究建议。请注意，下文讨论的几乎所有主题都适用于工程开发和社交互动实验。

First, in the topic of UAV form and appearance design, we recommend two kind of platforms in the engineering or social experiments: (i) a palm-sized multirotor UAV with cage design (e.g., 1]) for agile and accurate motion, safer interaction (less impact and less chance to get hurt by the propellers), affordance that invites touch interaction (e.g., 2]), if noise and flight time are not a big issue; (ii) a hugging-sized blimp with hiding propellers (e.g., 98) or novel propulsion unit with no rotating part (e.g., 116) for quieter and calm interaction, safer interaction, and longer interaction time, if agile and accurate response are not a big issue.
首先，在无人机的外形设计方面，我们建议在工程或社会实验中使用两种平台：(i) 手掌大小的多旋翼无人飞行器，采用笼式设计（如 1]），运动灵活准确，交互更安全（螺旋桨的冲击力更小，受伤的几率更小），可进行触控交互（如 2]）、2]），如果噪音和飞行时间不是大问题的话；(ii) 带有隐藏式螺旋桨的拥抱型飞艇（如 98 号飞艇）或无旋转部件的新型推进装置（如 116 号飞艇），如果敏捷和准确的反应不是大问题的话，可实现更安静、更平和的交互、更安全的交互和更长的交互时间。

Second, we suggest to pay more attention to integrated human-accompanying models and human-sensing interfaces in order to support a more realistic HDI. Human-accompanying model should integrate functions
其次，我们建议更加关注综合的人类陪伴模型和人类感应界面，以支持更加真实的人类发展指数。人类陪伴模型应整合以下功能

Fig. 3 Towards companion UAV from the autonomous UAV (left) and social UAV (right) categories.
图 3 从自主无人机（左）和社交无人机（右）两类中走向陪伴型无人机。

of human approaching, following, leading, side-by-side walking, bird-eye viewing on the top for a more natural HDI. Similarly, human-sensing interface should integrate at least four modalities of human tracking, hand tracking, face tracking, and voice interaction (e.g., 70 ). In the engineering field, UAVs should also perform environment sensing at the same time so that they can accompany their users without hitting obstacle (e.g., 65]).
为使人类发展指数更加自然，还应在顶部显示人类接近、跟随、引领、并排行走、鸟瞰等模式。同样，人感界面至少应整合人体跟踪、手部跟踪、面部跟踪和语音交互四种模式（例如 70）。在工程领域，无人机还应同时进行环境感知，以便在不撞击障碍物的情况下陪伴用户（例如，65]）。

Third, more related to the social interaction studies, we recommend to explore the ideas of sound design, tactile interaction, and holonomic flight motions of UAV. When individual interaction becomes more mature, we should try integrating the visual and audio expression, gesture and physical and tactile interactions, and investigate the long term HDI.
第三，在社会交互研究方面，我们建议探索无人机的声音设计、触觉交互和整体飞行运动的思路。当个体交互变得更加成熟时，我们应该尝试将视听表达、手势和身体触觉交互结合起来，研究长期的人类发展指数。

Fourth, we also encourage HDI researchers to share development code among companion UAV studies to facilitate comparison study under a shared metrics. In addition to our recommendation, one could also draw inspiration from the practices and know-how in the aerospace field 44 and social robotics field 20,37.
第四，我们还鼓励人类发展倡议研究人员在配套的无人机研究中共享开发代码，以便在共享指标下进行比较研究。除我们的建议外，还可以从航空航天领域 44 和社会机器人领域 20、37 的实践和技术诀窍中汲取灵感。

Fifth, last but not least, we suggest engineering research to (i) incorporate findings in the HDI studies (such as accompanying a user with a proximate distance that is comfortable to the user) into their technical development, and (ii) perform HDI study after developing a new function in order to confirm its usefulness and social acceptance (efforts from autonomous UAV to companion UAV, corresponding to the Fig. 3 (left)). At the same time, we suggest HDI studies to perform experiments with a real drone integrated with autonomous capabilities (such as accompanying a person and avoid obstacles autonomously) in order to deal with HDI studies in a more realistic scenario (efforts from social UAV to companion UAV, corresponding to the Fig. 3 (right)).
第五，最后但并非最不重要的一点是，我们建议工程研究(i)将人类发展指数研究中的发现（例如以用户感到舒适的近距离陪伴用户）纳入其技术开发中，以及(ii)在开发新功能后进行人类发展指数研究，以确认其实用性和社会接受度（从自主无人机到陪伴无人机的努力，对应图3（左））。同时，我们建议人类发展指数研究使用集成了自主功能（如陪伴人员和自主避障）的真实无人机进行实验，以便在更真实的场景中进行人类发展指数研究（从社会无人机到陪伴无人机的努力，对应图 3（右））。

9 Conclusion 9 结论

Technological advancements in small-scale UAVs have led to a new class of companion robots - a companion UAV. After identifying and coordinating related works
小型无人驾驶飞行器的技术进步催生了一类新的陪伴机器人--陪伴型无人驾驶飞行器。在确定并协调了相关工作之后
from the UAV's autonomy and sociability perspectives, we found that recent research efforts in companion UAVs focus on a single issue, either an engineering or a social interaction issue, rather than having an ambitious goal to tackle both issues in a paper. While this might be the nature of research (i.e., specialize on a topic), we encourage future works to emphasis on both aspects of companion UAVs as the integration of these two interrelated aspects is essential for an effective HDI.
从无人机的自主性和社交性角度来看，我们发现最近在伴航无人机方面的研究工作都集中在一个单一的问题上，要么是工程问题，要么是社交互动问题，而不是雄心勃勃地在一篇论文中同时解决这两个问题。虽然这可能是研究的本质（即专注于某一主题），但我们鼓励未来的研究工作强调伴航无人机的两个方面，因为这两个相互关联的方面的整合对于有效的人类发展倡议至关重要。

We also list up our observation throughout this review and propose guidelines to perform companion UAV designs and researches in the future. In addition to individual topics such as engineering functions and social interaction studies with new modality, we argue the importance of devising an integrated human-accompanying model and an integrated human-sensing interface to advance the development of companion UAVs. We also suggest researchers to share the programming codes used in their experiments to facilitate comparison study and consolidate findings in companion UAV works.
我们还列举了我们在本综述中的观察结果，并提出了未来进行陪伴式无人机设计和研究的指导原则。除了工程功能和新模式下的社会互动研究等个别主题外，我们还论证了设计综合人类陪伴模型和综合人类传感界面对于推动陪伴式无人机发展的重要性。我们还建议研究人员分享他们在实验中使用的编程代码，以便于进行比较研究和巩固伴行无人机作品的研究成果。

Most of the related papers focus on the development of the UAV itself. In contrast, it is also feasible to design an environment that could enable an easier navigation of companion UAVs. Public issues 90 and policy making 108 are also important for facilitating a social acceptance of UAVs. Lastly, while an affective model is an important aspect of companion UAVs, we argue that it is beyond the scope of this paper. We believe that affective models developed for general companion robots are applicable to companion UAVs when companion UAVs have more mature and integrated autonomous functions and socially interactive capabilities.
大多数相关论文都侧重于无人飞行器本身的开发。与此相反，设计一个能让无人飞行器更容易导航的环境也是可行的。公共问题 90 和政策制定 108 对于促进社会接受无人机也很重要。最后，虽然情感模型是伴航无人机的一个重要方面，但我们认为这超出了本文的范围。我们认为，当陪伴型无人机具有更成熟的综合自主功能和社会互动能力时，为一般陪伴型机器人开发的情感模型也适用于陪伴型无人机。

Compliance with Ethical Standards
遵守道德标准

Conflict of interest The authors declare that they have no conflict of interest.
利益冲突作者声明他们没有利益冲突。

The UAV database is shared online via Google Sheets https: //tinyurl.com/drone-paper-list). In the tables (one table per year), we list all the UAV-related papers from top journals/conferences since 2001 along with their related topics, abstracts, and details such as hardware summary. This list is particularly useful to: (i) search related works on a particular topic in UAV, e.g., HRI; (ii) search related works on a particular type in UAV, e.g., blimp; (iii) search related works on a particular UAV platform, etc. We believe that this list is not only beneficial for newcomers to the UAV field but also convenient for experienced researchers to cite and compare related works.
无人机数据库通过 Google Sheets https：//tinyurl.com/drone-paper-list在线共享。）在这些表格中（每年一个表格），我们列出了自 2001 年以来顶级期刊/会议上所有与无人机相关的论文及其相关主题、摘要和细节（如硬件摘要）。这份列表尤其有助于(i) 搜索无人机领域某一特定主题的相关作品，如人力资源创新；(ii) 搜索无人机领域某一特定类型的相关作品，如飞艇；(iii) 搜索某一特定无人机平台的相关作品等。我们相信，这份清单不仅有利于无人机领域的新手，也方便了有经验的研究人员引用和比较相关著作。

In addition to Google Sheets, we also use an open-source file tagging and organization software 105. TagSpaces enables readers to search papers with multiple tags or/and keywords effectively. Moreover, since original papers (PDF files) cannot be shared with readers due to copyright issues, for each paper entry, we create an HTML file that contains public information (such as abstract, keywords, country, paper URL link, and video URL link) for easier reference. To setup TagSpaces and download all the HTML files, please refer to our website at https://sites.google.com/view/drone-survey
除了 Google Sheets，我们还使用了一款开源文件标记和组织软件 105。通过 TagSpaces，读者可以有效地使用多个标签或/和关键词搜索论文。此外，由于版权问题，原始论文（PDF 文件）无法与读者共享，因此我们为每篇论文条目创建了一个 HTML 文件，其中包含公开信息（如摘要、关键词、国家、论文 URL 链接和视频 URL 链接），以方便读者参考。要设置 TagSpaces 和下载所有 HTML 文件，请访问我们的网站 https://sites.google.com/view/drone-survey 。

References 参考资料

Abtahi, P., Landry, B., Yang, J.J., Pavone, M., Follmer, S., Landay, J.A.: Beyond the force: Using quadcopters to appropriate objects and the environment for haptics in virtual reality. In: Proc. ACM Conference on Human Factors in Computing Systems (CHI), pp. 1-13 (2019)
Abtahi, P., Landry, B., Yang, J.J., Pavone, M., Follmer, S., Landay, J.A.: Beyond the force：使用四旋翼飞行器在虚拟现实中对物体和环境进行触觉控制。In：Proc.ACM Conference on Human Factors in Computing Systems (CHI), pp.
Abtahi, P., Zhao, D.Y., E., J.L., Landay, J.A.: Drone near me: Exploring touch-based human-drone interaction. Proc. ACM Interactive, Mobile, Wearable and Ubiquitous Technologies (IMWUT) 1(3), 813-824 (2017)
Abtahi, P., Zhao, D.Y., E., J.L., Landay, J.A.: Drone near me：探索基于触摸的人机交互。Proc.ACM Interactive, Mobile, Wearable and Ubiquitous Technologies (IMWUT) 1(3), 813-824 (2017)
Agrawal, H., won Leigh, S., Maes, P.: L'evolved: Autonomous and ubiquitous utilities as smart agents. In: Proc. ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp), pp. 487-491 (2015)
Agrawal, H., won Leigh, S., Maes, P.：L'evolved：作为智能代理的自主和无处不在的公用事业。In：Proc.ACM 泛在与无所不在计算国际联合会议（UbiComp），第 487-491 页（2015 年）
Aibot X6 (2016). URL https://www.aibotix.com
Aibot X6（2016 年）。URL https://www.aibotix.com
Arroyo, D., Lucho, C., Roncal, J., Cuellar, F.: Daedalus: A sUAV for social environments. In: Proc. ACM/IEEE International Conference on Human-Robot Interaction (HRI), p. 99 (2014)
Arroyo, D., Lucho, C., Roncal, J., Cuellar, F.: Daedalus: A sUAV for social environments.In：Proc.ACM/IEEE 人机交互（HRI）国际会议，第 99 页（2014 年）
Baytaş, M.A., Çay, D., Zhang, Y., Obaid, M., Yantaç, A.E., Fjeld, M.: The design of social drones: A review of studies on autonomous flyers in inhabited environments. In: Proc. ACM Conference on Human Factors in Computing Systems (CHI), pp. 1-13 (2019)
Baytaş, M.A., Çay, D., Zhang, Y., Obaid, M., Yantaç, A.E., Fjeld, M.: The design of social drones：关于居住环境中自主飞行器的研究综述。In：Proc.ACM Conference on Human Factors in Computing Systems (CHI), pp.
Bentz, W., Dhanjal, S., Panagou, D.: Unsupervised learning of assistive camera views by an aerial co-robot in augmented reality multitasking environments. In: Proc. IEEE International Conference on Robotics and Automation (ICRA), pp. 3003-3009 (2019)
Bentz，W.、Dhanjal，S.、Panagou，D.：增强现实多任务环境中空中协作机器人对辅助摄像头视图的无监督学习。In：Proc. IEEE International Conference on Robotics and Automation (ICRA), pp.
Bodin, T., Albin, M., Ardö, J., Stroh, E., Östergren, P., Björk, J.: Road traffic noise and hyper tension: Results from a cross-sectional public health survey in southern Sweden. Environmental Health 8 (2009)
Bodin, T., Albin, M., Ardö, J., Stroh, E., Östergren, P., Björk, J.：道路交通噪声与过度紧张：瑞典南部横断面公共健康调查的结果。环境健康 8 (2009)
Briod, A., Kornatowski, P., Klaptocz, A., Garnier, A., Pagnamenta, M., Zufferey, J.C., Floreano, D.: Contactbased navigation for an autonomous flying robot. In: Proc. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 3987-3992 (2013)
Briod, A., Kornatowski, P., Klaptocz, A., Garnier, A., Pagnamenta, M., Zufferey, J.C., Floreano, D.：基于接触的自主飞行机器人导航。In：Proc. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp.
Briod, A., Kornatowski, P., Zufferey, J., Floreano, D.: A collisionresilient flying robot. Journal of Field Robotics
Briod, A., Kornatowski, P., Zufferey, J., Floreano, D.：抗碰撞飞行机器人。野外机器人学杂志
Brock, A.M., Chatain, J., Park, M., Fang, T., Hachet, M., Landay, J.A., Cauchard, J.R.: Flymap: Interacting with maps projected from a drone. In: Proc. ACM International Symposium on Pervasive Displays (PerDis), pp. 1-9 (2018)
Brock, A.M., Chatain, J., Park, M., Fang, T., Hachet, M., Landay, J.A., Cauchard, J.R.: Flymap：与无人机投射的地图互动。In：Proc.ACM 普及型显示器国际研讨会（PerDis），第 1-9 页（2018 年）
Bruce, J., Perron, J., Vaughani, R.: Readyaimfly! Handsfree face-based HRI for 3D trajectory control of UAVs. In: Proc. Conference on Computer and Robot Vision (CRV), pp. 307-313 (2017)
Bruce, J., Perron, J., Vaughani, R.: Readyaimfly！用于无人飞行器三维轨迹控制的基于面部的免提人机交互技术。In：Proc.计算机与机器人视觉会议（CRV），第 307-313 页（2017 年）
Burke, M., Lasenby, J.: Pantomimic gestures for humanrobot interaction. IEEE Transactions on Robotics (TRO) 31(5), 1225-1237 (2015)
Burke, M., Lasenby, J.: Pantomimic gestures for humanrobot interaction.电气和电子工程师学会机器人技术论文集（TRO）31（5），1225-1237（2015）
Butzke, J., Dornbush, A., Likhachev, M.: 3-D exploration with an air-ground robotic system. In: Proc. IEEE/RSJ
Butzke, J., Dornbush, A., Likhachev, M.: 利用空地机器人系统进行三维探索。In：Proc.

Chun Fui Liew

The University of Tokyo, Japan
日本东京大学

E-mail: liew@g.ecc.u-tokyo.ac.jp
电子邮件：liew@g.ecc.u-tokyo.ac.jp

Takehisa Yairi 八里武久

The University of Tokyo, Japan
日本东京大学

E-mail: yairi@g.ecc.u-tokyo.ac.jp
电子邮件：yairi@g.ecc.u-tokyo.ac.jp
1 Details on the paper category analysis can be found in our previous survey paper 63. Details about the online sharing and regular updates can be found in Appendix A
1 有关论文类别分析的详情，请参阅我们之前的调查论文 63。有关在线共享和定期更新的详情，请参阅附录 A。
2 A common experiment setting used by researchers, where participants interact with a robot that participants believe to be autonomous, but in fact it is being manually controlled by a human behind the scene.
2 一种研究人员常用的实验设置，即参与者与一个机器人互动，参与者认为该机器人是自主的，但实际上它是由人类在幕后手动控制的。
5 Papers can be found by searching the "holonomic" keyword in the UAV paper list mentioned in Appendix A
5 在附录 A 中提到的无人机论文列表中搜索 "holonomic "关键词，即可找到相关论文。
We focus on thermal and touch interactions here, which have subtle difference with physical interaction (involves force feedback) mentioned in Section 6.4
我们在此重点讨论热交互和触摸交互，它们与第 6.4 节中提到的物理交互（涉及力反馈）有着微妙的区别

Companion Unmanned Aerial Vehicles: A Survey 配套无人驾驶飞行器：调查

Abstract 摘要

1 Introduction 1 引言

2 UAV Background 2 无人驾驶飞行器背景

2.1 UAV Definition 2.1 无人机定义

2.2 UAV Types 2.2 无人机类型

3 UAV Paper Identification Process3 无人机纸张识别过程

4 Perceptual Map of UAVs4 无人飞行器感知图

4.1 The Four UAV Categories4.1 四类无人机

4.2 Link with Previous Study4.2 与以往研究的联系

4.3 Research Efforts Towards Companion UAVs4.3 针对配套无人机的研究工作

5 From Remote-Controlled to Autonomous UAVs5 从遥控无人机到自主无人机

5.1 Human Following UAVs 5.1 人类跟踪无人飞行器

5.2 User Interfaces for UAVs5.2 无人飞行器的用户界面

6 From Remote-Controlled to Social UAVs6 从遥控无人机到社交无人机

6.1 Social Perception of UAVs6.1 社会对无人飞行器的看法

6.2 Emotion and Intention Expression of UAVs6.2 无人飞行器的情感和意图表达

6.3 Gesture Interaction with UAVs6.3 与无人飞行器的手势交互

6.4 Physical Interaction with UAVs6.4 与无人飞行器的物理交互

7 Observation and Lessons Learned7 意见和经验教训

7.1 UAV Form 7.1 无人飞行器形式

7.2 Appearance Design 7.2 外观设计

7.3 Human Accompanying Model7.3 人类陪伴模式

7.4 Human Sensing Interface7.4 人体感应界面