虚拟现实技术在绿色基础设施健康效益评估中的应用

著：（美）马修·布朗宁 （泰）彭瑟刚·萨布卡特派桑 姜珊 （美）安贾利·约瑟夫 译：翁羽西 袁帅

0 引言

研究表明，基因对人类健康与寿命变化的影响约占30%[1]。而影响健康的主要因素包括行为习惯（体育锻炼、睡眠和饮食状况）、医疗保健和物理环境。因此，随着城市化进程加快，越来越多的人涌入城市，人们的健康问题也愈发凸显。虽然逐渐完善的城市医疗水平能够为城市居民提供更好的健康保障，但城市居民同样也面临着由空气污染、生活压力等给健康带来的负面影响[2]。庆幸的是，城市规划、风景园林、公共卫生政策等领域的研究结果能够为降低城市化带来的负面影响提供一定参考，促进城市绿色基础设施的生态服务功能是改善人类健康的重要途径。绿色基础设施是指“能够保护自然生态系统的功能与价值，并由此为人类提供相关利益的、互相关联的绿色空间网络”，在城市规划与设计的背景下，植物也是低影响开发的重要构成要素[3-4]。具体来说，绿色基础设施元素包括人们所熟悉的行道树、绿化隔离带、花园和公园等，还有让人感觉“新奇”的元素，如绿色屋顶、植物墙、雨水花园、生态滤沟和湿地等[4]（图1）。在这2种情况下，绿色基础设施为生态系统带来了诸多益处，如城市雨洪管理、调节极端温度、为本地和跨地区物种（如鸟类和授粉昆虫）提供栖息地等[5-6]。

包括绿色基础设施在内的自然环境对人类健康的影响包括3个方面：降低城市化带来的弊端、恢复注意力、塑造环境的容纳能力[7]。植物可以通过吸附空气中的污染物、降低噪声的负面感知和隔离夜间的人造光来减少对人体的伤害[8-10]。自然环境还可以帮助我们恢复注意力，从人类进化角度看，自然界中熟悉的事物能够让人们获得更多的安全感和依恋感[11-15]。此外，绿色空间不仅为促进公众健身活动、社会交往提供机会，还有益于改善睡眠，维持人体肠道、皮肤等部位的微生物群落平衡[16-18]。

大量的科学实验和观察性的研究证实绿色基础设施有益于促进人类的健康和福祉。近期的文献总结回顾了绿色基础设施在某些疾病中所发挥的健康效益，包括哮喘[19-20]、心血管健康[21]、老年痴呆[22-23]，以及在生育健康[24]、心理健康[25-26]、死亡率[27]、压力[28]等各个方面的益处。

不同类别的绿色基础设施可能发挥不同的健康效益[4]。绿色基础设施中为居民所熟悉的景观要素可以发挥保护性的效益，这一点已经有广泛的研究支撑[4,29-30]。然而另有小部分的研究表明，诸如生态植物墙、绿色屋顶、生物滞留系统，以及其他并不为人所熟知的绿色基础设施也能发挥一定的健康效益[7,27,31-33]。在绿色基础设施的植物类型方面，有观测研究发现，相较于单纯的草本植被，人们住宅区附近的乔木覆盖程度与人体的健康体质指数（BMI）水平密切相关[34]，并且影响居民在医院外的死亡率[35]以及小学生的学习成绩[36-37]等。相关的VR实验也支持以上的结论。相比草本植物景观，居民区附近的乔木和行道树景观可以使人更快地从应激性压力中恢复过来，这些是通过测量焦虑、压力和回避性行为的调查问卷[38]，以及被试者唾液中更低的皮质醇指标等体现出来的[39]。除此之外，通过VR营造的人造森林里，如果座椅区域附近空旷、视野良好，则比隐藏在密林中的座椅有更强的压力恢复效益[40]。

影响绿色基础设施健康效益的其中一个因素是其被感知的安全性。超过45项的研究发现，多种类型的植物空间会引发人们的恐惧和焦虑感[41]，而恐惧感会妨碍人们在使用这些公共空间时获得健康方面的益处[42]，并且对人们的主观幸福感产生消极的影响[43]。容易让人产生恐惧感的空间因素包括：及膝至视平线高度范围内植物密度过高，杂乱的灌木丛和封闭的围合空间，例如道路两侧的植被或广场四周[41]。植被的空间布局（可见边界）和通透性（深度、高度、孔隙度）共同影响感知的安全性和主观恢复性[44]。

景观偏好也会影响绿色基础设施的健康效益。与偏好度低的自然环境或室内环境相比，当VR用户在观看他们喜欢的自然环境（如海滩、田园景观）时，悲伤、不安、易怒和紧张等负面情绪显著降低[45]。数十项研究都表明，绿色空间的健康效益也同样受到景观偏好的影响[33,46]。

由于影响绿色基础设施健康效益的因素诸多，因此，针对绿色基础设施要素的循证研究能够为具体的环境设计提供有价值的参考。然而，依然存在许多局限性。例如，在实验室中开展研究可能会对心理和生理指标造成一定影响，天气条件、交通等因素也会使得实景实验可操作性降低，而且采用2D图片充当感知媒介可能会导致受试者对实景中的感知敏感性下降[47-48]。

VR技术的新手段，提高了研究的生态效度，为绿色基础设施健康效益研究的开展提供了可能。生态效度是指在实验条件下参与者的行为和感知与真实现象相符的程度[49-50]。

VR技术可以作为人类健康治疗的一种替代手段在过去多项研究中得到证实。Nukarinen等发现，在利用VR技术模拟的360°立体森林视频中暴露10 min能够有效地降低交感神经系统活性（即“类似‘战斗或逃跑’的过度应激反应”），抑制负面情绪。与真实森林环境相比，虚拟环境更有益于主观情绪的恢复[51]。Browning等的研究表明，在真实的森林里停留6 min，对受试者在情绪调节与注意力恢复上的效果与观看360°森林视频相当，而这两者（真实与虚拟的森林环境）均比没有绿色基础设施的对照组有更显著的健康效益[52]。Chirico和Gaggoli的研究也证实，暴露在真实和虚拟的全景式山湖景观场景所减少的消极影响和引起敬畏感的程度相似[53]。Yin等利用VR技术观察4种不同的办公室环境（“亲生物设计”与“非亲生物设计”）对受试者压力和焦虑恢复的影响。结果发现，在“亲生物设计”的办公室环境中停留5 min，与观看360°“亲生物环境”视频产生的健康效益（血压和心率）相似[54]。

本研究概述了VR系统、内容创作、实验设计、健康指标测量和安全注意事项等多个重要问题，为研究人员开展相关实验提供了科学的方法建议和指导。

1 VR系统

VR系统主要分为2类：一是沉浸式物理空间，如洞穴式自动虚拟环境（CAVE），这类系统是房间大小的立方体空间，带有视频投影仪，可将移动的图像投影至参与者周围的半透明屏幕；二是头戴式显示设备（HMD）系统，涵盖了用于展示视听刺激的护目镜。下文将简要介绍这2类系统的历史和使用情况。

1.1 沉浸式物理空间

CAVE由伊利诺伊大学香槟分校的电子可视化实验室发明，并于1992年首次向广大观众展示[55]。CAVE系统通常包括一个3.33 m3的立方体房间，光线较暗。房间内4～6个侧面都配有投影幕。以一定角度旋转的反射镜放置在短焦投影仪和投影幕之间，可将高分辨率场景投射到投影幕上[56]。

CAVE可有效操纵触觉信息，因而其沉浸感、真实感和体验感比其他形式的VR更强[57]。该系统可有效激发与环境设计存在关联的情绪，例如与恐高症相关的焦虑和恐惧[58]，并能辅助寻路决策和导航[59]。然而，CAVE系统需要昂贵的初始安装费用和高级的计算机技能。该系统还要求较大的实验室空间，其可移植性和灵活性都十分有限。或许恰恰因为空间需求大、成本高、要求技术专长[60]，以及对设备的设置和调试有很高的要求[56]，CAVE系统很少用于对绿色基础设施的健康益处的研究（存在2个例外）[61-62]。

1.2 头戴式显示设备

基于HMD的VR发轫于20世纪60年代[63]，但对于研究者，此类设备直至近期才被引入到科研领域中[64]。2012年，Oculus Rift的问世标志着第二次HMD开发浪潮，推广了低价、舒适且高质量的设备。新设备、新软件百花齐放。HMD VR已有不少于50种型号，每年都有新型号问世，旧型号也有大量更新。至少有2篇科学文献介绍了可用的HMD型号及其提供沉浸感的潜力（分辨率、帧速率和视野），对研究的效用（可移动性和成本）以及用户体验[65-66]。由于HMD高速发展，最新信息的来源可能是VR专业组织网站，如Virtual Reality Society①。截至本文撰写时，HMD可分为三大类：手机HMD、桌面HMD和一体机HMD（图2）。

手机HMD是价格低廉、便携性好且易于使用的头戴设备，需要使用智能手机投影图像、播放声音。Google Cardboard是一个较为知名的例子，而其他型号的设备有不同的沉浸体验[67]。 Oculus Gear VR和Google Expedition套件对于研究者较为陌生，但它们质量高，并且价格仅相当于一体机和桌面机的一小部分[66]。然而相较之下，手机HMD受限于手机运算能力，用户体验和控制能力都较差。并且播放高分辨率和高帧率影像需要高端手机，后者的价格或与其他类型HMD相当，甚至更高。手机HMD售价通常为70～350人民币（约10～50美元），不包含手机价格。

与手机HMD不同，一体机HMD内置了处理器、GPU、显示器、内存、电池和传感器，因此不需要连接其他设备。较知名的型号包括Oculus Go和Oculus Quest。一体机HMD不受线缆束缚，某些型号更内置侦测物体的传感器，可防止用户意外触碰周边物体。一体机HMD的体验比手机HMD更真实，但体验的复杂性无法与桌面HMD相比。一体机不同型号的价格在1 400～3 500人民币（约200～500美元）之间。

桌面HMD是最高级机型，需要配合计算机使用，它配备遥控器、头部跟踪器和收集外部数据的传感器。这一类型包含Oculus Rift和HTC VIVE。桌面HMD还使实验研究易于控制和操纵，提高生态效度，实现完全沉浸和高度真实感。桌面HMD的初始设置需要操作者具备较高的计算机技能，因为计算机软件不如其他类型HMD软件对初学者友好。桌面系统的价格最为昂贵，其HMD约为7 000元人民币（约1 000美元），它还需要功能强大的台式计算机和专用显卡来提供流畅的渲染。

2 VR内容创作

虚拟环境可以完全由计算机生成，也可以基于由360º全景视频记录并且经过编辑的真实环境。虽然这些方法专注于视觉输入，但声音（比如流水潺潺、树叶沙沙作响）可能同样对绿色基础设施的健康效应至关重要[61]。自然环境中的气味也会诱发有益的生理变化[68]。因此，VR体验的多感官特点为前沿研究者提供了大量的研究机会。但笔者暂不对多感官体验做进一步讨论。基于我们对相关文献的批判性回顾，以及我们的科研项目经验，我们确信那些对绿色基础设施的健康效应有兴趣但不熟悉VR技术的入门级研究者，至少在项目的初始阶段，应该把研究重点放在人类最主要的感官——视觉领域里[69]。另一点原因是，我们对多感官输入与情绪反应的关系还知之甚少[70]。因此，在视觉、声音和气味等感官体验相互影响的实证证据出现前，基于视觉的多场景比较或许比多感官比较具有更高的生态效度[70]。

2.1 计算机生成的环境

不同的VR系统需要不同的程序来创建和显示绿色基础设施。对于纯粹由计算机生成的模型和动画，首先需要在游戏引擎（即Unity 3D、Autodesk 3ds Max或Maya）或专业建模程序（即SketchUp或Autodesk Revit）中生成内容。游戏引擎具有内置函数，可供在单个程序中建模、渲染、生成VR环境。游戏引擎为创建可交互的环境提供了强大支持，它可以模拟多种场景，提供导航和交互选项。环境影响评估相关研究中，也存在使用游戏引擎在VR中创建绿色基础设施内容的早期尝试。譬如，Gang、Choi、Kim和Choung等[71]开发了使用Unity 3D和WebGIS系统的工具套件，可在VR中显示水文和水灾信息。其他研究介绍了使用游戏引擎（Unity 3D）创建VR内容的基本概念和过程[72]。

对于刚入门的VR研究者，创建计算机生成环境的第2种途径可能更容易。这种方法从BIM建模开始，然后将BIM模型转换为VR环境。通过移动应用（InsiteVR或Kubity）或台式机的应用程序（eyecad VR或Enscape）得到实际的VR体验。移动应用和计算机应用平台的主要区别在于，前者需转换整个VR模型，后者支持实时渲染。这种方法在近期一项（虚拟的）在医院花园进行的空间认知和寻路研究中得到了应用[73]。图3展示了VR用户所看到的医院花园实时渲染品质。

在VR中创建、展示绿色基础设施内容时，研究者也可以混合上述2种工作流。如Yu、Behm、Bill和Kang[74]使用Autodesk 3ds Max和Unity 3D研究了周边风电厂声景在视觉和噪声影响方面的差异。Lin、Homma和Iki[75]的一项相似研究使用ArcGIS 10.0和Adobe Photoshop CS6生成3D模型，研究了人对不同尺度蓝色空间（水景）的视觉偏好。

2.2 计算机修改的真实环境

生成VR内容的另一种方法则截然不同。这种方法是摄录实景，再进行编辑。实景可由特殊的相机录制，该设备附带不少于2枚180º视角鱼眼镜头。随后，将各镜头录制所得画面拼接合一，并保持时间同步。可以使用视频或照片编辑工具（如Adobe Photoshop、Premiere或After Effects）对拼接的视频进行编辑，从而添加、删除或编辑绿色基础设施要素（树木、草丛、建筑物或公园长凳）。具备编辑二维图像、视频经验的建筑环境研究者或许对该工作流程更为熟悉[44]。要获得多种绿色基础设施场景，也可以直接拍摄多条视频，并进行简单的后期处理。例如，Hedblom等[68]的后期处理仅仅包含去除三脚架、锐化图像，使用Adobe Photoshop校正光线。Calogiuri等[54]使用 Adobe After Effects、Warp Stabilizer VFX和Samsung Gear 360 ActionDirector来改善视频稳定性。此外，Browning等[52]称其仅对视频进行了轻微的光线调整。图4提供了有关如何编辑360º全景视频从而改善图像，修改绿色基础设施要素的示例。

3 VR实验设计

实验性研究是研究人员用于寻找对绿色基础设施健康效益影响较大的关键因素的有效方法。不同于相关研究或准实验研究，实验性研究通过实验手段施加于实验客体的某些干预因素，观察由此引起的实验效应。在没有随机分组和对照组的情况下，很难判断特定的绿色基础设施是否会导致受试者的心理效益产生变化。换言之，许多类型的研究都可以使用VR实验设计来评估其对健康的影响[76]。

研究绿色基础设施健康效益的实验方法通常分为被试间设计和被试内设计。以下将对实验设计方法、造成实验误差的因素以及如何避免误差项展开介绍。

3.1 被试间设计

被试间设计是指将被试者随机分配在不同组接受不同的自变量处理。除了自变量外，每组被试者都接受相同的实验处理，即每个人只接受一种处理。例如，在VR实验中，对照组感知的场景没有设置绿色基础设施，而实验组能观察到不同绿色基础设施的元素。实验必须遵循“随机法”来最大限度地消除与研究相关的干扰因素（如习惯、行为等）以保证实验结果的客观性。通常，实验进行简单的随机化，被试者有同等的概率被分配到A组、B组或C组等，而不受研究者或受试者的主观意愿影响。但在样本量较小的情况下，可以采用组块随机化，以确保各组参与人数相等[77]。心理学和临床医学的研究通常要求不同对比组实验对象在性别上保持一致，但在绿色基础设施健康效益研究中，其他特征因素同样重要，包括人群年龄、民族、职业、经济状况、对景观的偏好/熟悉程度、在环境中度过的时间、与自然的联系程度、被感知的安全性、VR体验的经验等[46]。

3.2 被试内设计

被试内设计是指被试者接受所有自变量水平处理的实验设计，也称为重复测量设计。这种实验设计能够使被试者成为自己的“对照组”，提升统计功效，同时能有效检测到弱效应（效应在实验中用来描述组间差异的大小）。这些优势使得被试内实验设计更受研究人员的青睐，尤其对于一些引入自然仿真模拟的实验，包括虚拟绿色基础设施[46]。

被试内设计也有缺点：1）被试者必须在实验中花费更多的时间，根据实验设计，被试者可能需要长期、多次地访问实验室，这将导致被试者流失，从而减少样本量；2）顺序效应也会干扰实验结果，例如，参与者多次练习造成“学习效应”，或者在漫长的实验过程中感到疲惫，从而影响主观问卷结果。

为了减小实验误差可以平衡执行实验条件的顺序，例如，1/2的参与者可以先看场景A，然后再看场景B，另外1/2的参与者则以相反的顺序观看。让参与者在实验开始前练习问卷或认知任务也可以平衡“学习效应”[78]。

3.3 实验有效性的影响因子

实验设计往往受几种类型偏差的影响。需求特征是指被试者根据主试者的暗示和期望有倾向性地调整自己的行为或回答问卷。这种偏差是被试内实验设计特别关注的问题，因为被试者经历了多种实验条件，很可能已经发觉了各实验条件之间的不同，从而猜测到了主试者的意图[79]。实验人员也会根据自己对实验结果的期望而无意识地调整自己的行为，这种偏差被称为预期效应。在不同的实验条件下，主试者可能会通过语言和肢体动作或表达或抑制自己的热情[80]。

这些误差可以通过双盲控制来减少，被试者和主试者都不知道实验的内容、目的、分组和实验条件，从而有效地消除可能出现的被试者和主试者意识当中的主观偏差和个人偏好，但这种实验方法在VR研究中很难实施。一种双盲控制方法是采用被试内设计，以随机的顺序呈现虚拟环境，每名被试者观看场景的顺序均不相同，软件同步记录每名参与者呈现的场景顺序。

4 健康测量

实证数据的积累，为模拟的绿色基础设施场景能够促进人类健康提供了有效依据。例如，观看VR场景中的植物墙和广阔的森林景观有益于缓解压力和减少焦虑[11]。研究人员运用VR技术来观察人的情绪变化，结果发现树木、草地和溪流比没有绿色基础设施的城市环境更能改善人的情绪状态[81]。与室内环境[52]、抽象画[82]或没有绿色基础设施的城市环境相比[83]，森林环境和开阔视野的草坪对人情绪放松或情感唤起的作用更明显。相较于只有建筑和树木为背景的景观环境，城市公园和森林能够更好地促进人们的积极情绪和注意力恢复[84]。多项研究已经证明了行道树的减压潜力[38-39]。环境影响研究中的测量方法主要有客观测量（生理响应）和主观测量（心理响应），这2种方法可以用来观测生理唤醒、主观情绪和认知表现的结果。

表1 用于量化VR中绿色基础设施导致的情绪健康/认知能力变化的工具和方法Tab. 1 Survey instruments and approaches to measuring self-reported changes in emotional health/cognitive performance resulting from GI in VR

4.1 生理测量

生理反馈仪主要用来测量由于环境刺激引起的生理唤醒表征，可以反映心血管系统，如血压（BP）、血容量脉搏（BVP）、心率（HR）和心率变异性（HRV）[85]，也可以反映内分泌系统，如唾液或血液中的皮质醇浓度。

心血管系统和内分泌系统被证实与人类长期的身体健康有关。心率升高会增加血栓、高血压、中风和其他心血管疾病的患病风险[86]。此外，也可以作为压力的客观指标[87]，慢性压力会导致一系列身心健康问题，如失眠、生殖健康问题或癌症[88]。

如果对生理反馈测量准确即可获得高质量的数据。而且随着实验设备成本越来越低，可穿戴式装置采集技术逐渐成熟，使得采集生理数据的难度降低[85]。然而，依然存在一些问题，比如在实验过程中传感器不灵敏导致的无效数据会影响结果的分析[89]。在一些应激激素采集的过程中，侵入式的采集方式也可能给参与者带来无形的压力，并需要延长实验时间。例如，血液中的皮质醇扩散进唾液中需要20 min，在采集过程中，受试者被要求留在实验室中[95]。一段时间内摄入的食物和咖啡因饮料，以及吸烟、体育活动、睡眠和性别因素也可能影响皮质醇水平和其他生理指标的波动[91-92]。想要获得有效的实验数据，必须严格遵守操作规程，以控制额外变量造成的实验混淆效应。

4.2 心理和认知功能测量

情绪变化、注意力和短时工作记忆是绿色基础设施影响心理健康和认知能力的主要方面[32,46]。心理健康效益的评价方法主要采用向被试者呈现某些特定情境，然后测量他们的生理指标，或使用一系列标准化的心理调查问卷采集数据。认知能力是心理健康的一种形式，包括信息的处理、做出决定和在生活中获得成功的能力[93]。

生理测量包括身体对压力或消极情绪的反应，主要通过面部肌肉运动，比如前额肌肉紧张导致的皱眉[94]、皮肤电导率[95-96]以及大脑活动来反映[97]。这种测量的优点在于不受被试者主观意识的影响，能够准确地反映被试者的客观情绪和生理变化，但有些会被其他身体信号所干扰，比如脑神经活动造成的认知负荷[98]。这些数据的分析难度大，而且采集分析脑电数据的成本更高。

心理量表是心理学研究及应用的重要工具，优点之一是研究人员易于操作。问卷测量通常在VR环境体验之前和之后进行，测试前和测试后的差异对比用于说明确定环境条件对某方面心理或行为变化的效应。常见的心理健康评价方法如表1所示。

评估注意力和短时工作记忆的实验设计通常先引入让被试者感到疲劳或有压力的任务，然后向被试者呈现一系列绿色场景后对认知任务的完成程度进行评估。这一研究假设基于环境心理学的“注意力恢复理论”，该理论认为长时间使用定向注意力会发生功能衰退，自然景观能够有效地将人的注意力从定向注意向非定向注意转移，从而缓解精神疲劳[99]，这是绿色基础设施健康效益的主要理论之一[13]。根据另一个心理进化理论“压力缓解理论”，压力引入环节也可以应用于其他研究内容，如被感知的压力[100-101]。这也解释了为什么绿色基础设施有益于人类的健康[102]。

现有的技术还有很多，但是并未在绿色基础设施的健康效益研究中获得广泛的应用。值得注意的是，眼动追踪技术可用于揭示受试者观看不同绿色基础设施时的眼动特征并进行分类[11]。HTC Vive Pro Eye VR眼镜搭载了眼球追踪系统，该设备通过多个指标（扫视和注视）记录360°场景或3D环境中的眼球中央凹关注点。借助生物传感器可以同步获得生理、心理数据例如皮肤导电性等，多种技术的融合帮助我们更好地了解受试者的视觉注意力被哪些绿色基础设施特征吸引，何种环境要素影响人们的情绪感受。

5 安全注意事项

进行VR场景实验的潜在风险较低，但依然有眩晕的可能性。此外，随着新冠肺炎的肆虐使得传染病防控也成为需要高度警惕的安全问题。

5.1 眩晕

眩晕是指感受自身或外界物体运动性幻觉的各种症状，VR实验中眩晕可由前庭系统刺激（物理运动）或视觉刺激（观察到的运动）引起[116]。症状包括眼睛疲劳、头痛、面色苍白、出汗、口干、胃胀、定向障碍、脑内摇晃、共济失调、恶心、呕吐、唾液分泌增多、打嗝等[116-117]。

眩晕的原因有很多，包括用户个体特征、VR环境中的运动、暴露时间和频率，以及设备规格（帧速率、视场和分辨率）[118]。有研究发现VR用户在虚拟空间中“身临其境”的感觉程度与眩晕症状呈负相关[119]。针对这些原因，解决方法包括使用高质量、舒适的HMD，减少三维环境中物体的视觉复杂性，例如增加三角形、顶点和纹理大小的数量，而不牺牲真实性来减少延迟和优化帧速率[76,120]。

5.2 传染性疾病传播

VR设备通常佩戴在人们的头部，靠近眼睛、鼻子和嘴，这为细菌、病毒和真菌在人与人之间传播提供了机会。在VR实验中对仪器进行事先清洁是十分必要的，不遵守实验管理规范可能会违反伦理委员会的规定和要求，导致实验被终止，HMD也可能成为疾病传播的媒介。HMD中的海绵可以吸收体液，比如汗液。一次性口罩和可替换的海绵垫可以在不同的被试者之间进行替换，两者大约是相同的金额（10～20美元/个，约70～140元/个）。

每次使用后，需要对HMD进行清洗消毒②。同时可以配合紫外线（UV）灯进行消毒杀菌。据报道，紫外线灯可以在60 s内杀死耳机表面99.99%的细菌、病毒和真菌③。

6 结论

VR技术为研究人员探索绿色基础设施健康效益提供了一种令人兴奋、方便且相对廉价的方法。VR实验必须仔细考虑研究设计、健康测量指标、安全预防措施、伦理道德审查等。由于这项技术的快速发展，研究人员可以将笔者提出的建议与其他VR研究的实际情况相结合，以确保设备的可操作性和实验的可实施性。面对日益严峻的城市化问题，绿色基础设施建设在促进健康型城市方面应发挥关键作用，将健康和健康风险数据应用到绿色公共空间规划，有利于实现全球共同利益最大化。

致谢：

感谢美国克莱姆森大学博士生鲁塔利·乔希和研究生尤尼则·赫拉曼提供的图像支持。

注释：

① https://www.vrs.org.uk。

② HMDs的安全使用的更多信息，建议浏览以下网站：https://aixr.org/press/articles/covid-19-safety-for-virtualaugmented-reality-aixr-guidelines/、https://uploadvr.com/sanitize-clean-vr-headsets-oculus/。

③ https://www.cleanboxtech.com。

图表来源：

图1由彭瑟刚·萨布卡特派桑提供；图2、3由姜珊提供；图4由鲁塔利·乔希和尤尼则·赫拉曼提供；表1由马修·布朗宁绘制。

Human Health Assessments of Green Infrastructure Designs Using Virtual Reality

Authors: (USA) Matthew Browning, (THA) Pongsakorn Suppakittpaisarn, JIANG Shan,(USA) Anjali Joseph Translators: WENG Yuxi, YUAN Shuai

0 Introduction

Only an estimated 30% of our health, wellbeing, and life span is determined by our genetics[1].Other major drivers of health are behavior (physical activity, sleep, and eating patterns), access to health care, and the physical environment. Thus,when people increasingly move to urbanized areas, their health is likely to change. While cities can improve health by improving access to high quality healthcare, they can also degrade health by negatively impacting some health-promoting behaviors and by exposing residents to possibly stress-inducing and toxic environments[2].

Fortunately, researchers studying urban planning, design, public health policy, and other aspects of the built environment can play a role in reducing the potential negative impacts of cities on health. Incorporating green infrastructure(GI) into cities can provide array of beneficial ecosystem services that improve human health.GI refers to the “interconnected network of green space that conserves natural ecosystem values and functions and provides associated benefits to human populations”[3], but in the context of city planning and design, GI also includes the small but interconnected vegetative elements in outdoor designs[4]. These elements can be quite familiar (i.e.,street trees, median plantings, gardens, and parks)or less unfamiliar or “novel” (i.e., green roofs, green walls, rain gardens, bioswales, and constructed wetlands) to city residents[4](Fig. 1). In both cases,GI provides an array for ecosystem benefits, such as managing urban stormwater, mitigating extreme temperature, and providing habitats for local and transregional species such as birds and pollinating insects[5-6].

The health-related impacts of any natural setting, including GI, encompass three domains:reducing harm, restoring capacities, and building capacities[7]. Vegetative elements can reduce harm by filtering air pollutants, buffering against noise,and reducing artificial light at night[8-10]. Vegetative elements can also restore our attention and focus through exposure to elements in the landscape that humans have evolved in, and therefore feel safe in, or have otherwise become familiar and preferred[11-15]. Finally, vegetative elements build the capacity for physical activity, sleep, social interaction, and commensal bacteria in the gut, the skin and elsewhere[16-18].

A large number of experimental and observational literature supports the consensus that GI can benefit people. Recent reviews have synthesized the available research for specific health outcomes, including asthma/atopy[19-20],cardiovascular health[21], dementia[22-23], birth outcomes[24], mental health[25-26], mortality[27], and stress[28].

The ability of GI to activate these pathways toward health may depend on the type of GI[4].A large body of evidence supports protective effects of GI elements with which many residents are familiar[4,29,30], while a much smaller mounting body of evidence is available for the protective effects of green walls, green roofs, bioswales, and other less-familiar elements[7,27,31-33]. Regarding types of vegetation, observational studies have found that tree canopy cover levels near people's homes are more strongly associated with healthy body mass index (BMI) levels[34], protection from out-of-hospital deaths[35]and school test scores in elementary schools[36-37]than herbaceous or grass cover. Additional support for these findings has been shown in experimental studies. Higher levels of residential street tree canopy cover in virtual environments have facilitated recovery from an acute stressor in measures of self-reported levels of anxiety, tension and avoidance scores[38]and salivary cortisol levels in men[39]more than lower levels of canopy cover. In addition, made-made elements in forested settings have been shown in VR to have differential effects on recovery from an acute stressor — seating areas with sufficient open space between the seats were more restorative than settings with more obscure seating areas[40].

Another mediator of the health benefits of GI are their perceived safety. Over 45 studies have observed that many types of vegetative space could invoke fear and threats to perceived personal safety[41], and fear can prohibit the health benefits of using these spaces[42]as well as negatively impact subjective well-being[43]. Factors that are more likely to induce fear include high levels of vegetative density between eye level and knee level, untidy clusters of shrubby vegetation, and complete enclosure (i.e., vegetation on both sides of a path, or all sides of a plaza/open space)[41].Both the spatial arrangement of vegetation (the boundaries of visibility set by vegetative planting)and permeability of vegetation (depth, height,and porosity) have been shown in VR to influence perceived safety and perceived restorativeness[44].

Landscape preferences may also influence the protective effects of green infrastructure.Negative moods (i.e., distress, upset, irritable, and nervous) were lowered more strongly when VR users watched a natural setting that they preferred(a beach or pastural agricultural setting) compared with a natural setting that they preferred less or an indoor setting[45]. Similar mediating effects of landscape preferences on health benefits of green spaces have been shown in dozens of observational studies[33,46].

Due to the many factors that influence the health benefits of GI, new empirical research with populations and GI elements of interest would best inform context-specific design interventions.Yet researching psychological and physiological indicators of health in field experiments or laboratory settings with two-dimensional (2D)imagery is inadequate; weather conditions and travel requirements make field experiments expensive and difficult to conduct, and participant reactions to non-immersive 2D imagery may not represent how they would respond in the real world[47-48].

Here enters virtual reality (VR) technology.VR provides the opportunity for researchers to test the health benefits of multiple environmental scenarios with high levels of ecological validity,which describes the extent to which participant's behaviors and perceptions in controlled research settings mimic the real-world[49-50].

This use of VR has been validated in several studies that compared responses to physical GI elements and their virtual counterparts. Nukarinen et al. found that a 10-minute forest exposure decreased sympathetic nervous system activity (the“fight or flight response”) and negative emotions similarly as exploring a three-dimensional (3D) or 360-degree forest video[51]. Browning et al., showed that improvements in positive emotions and the perceived restorativeness of a 6-minute real forest exposure was similar to watching a 360-degree forest video when each setting (real and virtual)was compared to an indoor setting without GI[52].Chirico & Gaggioli demonstrated that a 5-minute exposure to a panoramic mountain and lake view decreased negative affect and induced awe to a similar extent as a 360-degree video of the same scene[53]. Yin et al. compared environments with and without plants or views of GI (“biophilic”versus “non-biophilic” environments) and discovered that a 5-minute exposure to a biophilic indoor environment resulted in similar beneficial changes to blood pressure and heart rate as watching a 360-degree video of a biophilic environment[54].

This essay summarizes and makes recommendations for researchers interested in health promotion through the GI. We describe VR systems, content creation, study design, health outcome measurement, and safety considerations.Our objective is to support researchers in evaluating the impacts of GI elements on health outcomes.

1 VR Systems

There are two main types of VR systems.Immersive physical spaces, such as the Cave Automated Virtual Environment (CAVE), are room-sized cubical spaces with video projectors that direct moving imagery on translucent screens surrounding the participant. Head-mounted display (HMD) systems involve goggles worn on the head that display visual and acoustic stimuli.Brief reviews of the history and use of these two systems are provided below.

1.1 Immersive Physical Spaces

The CAVE was invented by a group of researchers at the University of Illinois at Urbana-Champaign in 1992[55]. Such systems generally include a 3.33 m3cubic room with darkened lighting conditions. Four to six sides of the room are equipped with projection screens. Scenes display on the screens are reflected by mirrors positioned and rotated between high-resolution,short-throw projectors and the screens[56].

The CAVE allows effective manipulation of tactile/haptic cues, which enhances immersion,realism, and experienced presence over other forms of simulated environments[57]. These systems have been shown to be particularly effective at provoking certain emotions related to environmental design, such as anxiety and fear associated with acrophobia[58], and facilitating way finding decisions and navigation[59]. CAVE systems have been rarely used in studies of the health benefits of exposure to GI, likely because of the large space requirements, high cost and technical expertise[60]and extensive setup demands[56]. However, there are at least two exceptions[61-62].

1.2 Head-Mounted Displays

Head-mounted VR displays were proposed in the 1960s[63]but have been unavailable to most researchers until recently[64]. In 2012, the introduction of the Oculus Rift signaled a second wave of HMD development with devices that were inexpensive, comfortable, and of high-quality.A vast range of new devices and software have since been developed. At least two reviews in the scientific literature describe the available HMD models in respect to their immersive potential(i.e., resolution, frame rate, and field of view),research utility (i.e., mobility and cost), and user experience[65-66]. Given the rapid development of HMDs, the websites of established VR professional organizations, such as the Virtual Reality Society①,may provide updated information. At the time of writing this article, there are three broad categories of HMDs: phone-based, tethered, and all-in-one(Fig. 2).

Phone-based HMDs are affordable, portable,and easy to use headsets that require a smartphone to project imagery and emit sound. The Google Cardboard is a well-known example, albeit unrepresentative of the immersiveness available in other models[67]. The Oculus Gear VR and Google Expedition kits may be less familiar to many researchers but are relatively high-quality devices that cost a fraction of the cost of other HMD systems (i.e., all-in-one and tethered)[66]. Due to the limited processing power of some smartphones,phone-based HMDs may deliver a lower-quality experience with reduced user controls compared to other types of HMDs. Also, premium smartphones are required to display high-resolution, high-frame rate imagery in these headsets, and the price of compatible smartphones can match or supersede the cost of other HMD options. For the headset without the accompanying phone, the price of a phone-based HMD generally ranges between 70 and 350 RMB （10 and 50 USD).

Unlike phone-based HMDs, all-in-one models include built-in processors, GPUs, displays,memory, batteries, and sensors, so no additional equipment is necessary. Notable examples of these devices are the Oculus Go and Oculus Quest. Such HMDs are wireless and sometimes have builtin sensors for detection of physical objects in a room to prevent the user from colliding into them.All-in-one HMDs can offer a more powerful VR experience than phone-based systems but may deliver a less sophisticated experience than tethered systems. All-in-one models generally range in price between 1 400 to 3 500 RMB (200 to 500 USD.)

The most sophisticated HMDs are tethered to external computers and come with remote controllers, a head tracker, and external data collection sensors. Examples include the Oculus Rift and HTC VIVE. Advantages of tethered HMDs include high experimental control and manipulation,high ecological validity, full immersion, and high realism. Tethered systems require a higher level of computer skills for the initial setup, because the computer software is often not as beginner friendly as software that accompanies other types of HMDs.Tethered systems are highest in price, around 7,000 RMB (1,000 USD) for the HMD, and require a powerful PC with dedicated a GPU to provide fluid rendering.

2 VR Content Creation

Virtual environments can be entirely computer generated or they can be based on real environments that are recorded in 360º videos and digitally edited. Although these approaches focus exclusively on visual inputs of virtual environments, sounds (i.e., water flowing, leaves rustling in the wind) may be essential to the health-promoting effects of GI[61]. Similarly, many aspects of the natural environments emit smells that can induce beneficial physiological changes[68].Therefore, multisensory VR experiences provide rich opportunities for trendsetting researchers.Such experiences are not discussed further in the current essay, however. Our critical review of the literature and professional experience leads us to believe that researchers interested in the health benefits of exposure to GI but unfamiliar with VR will focus — at least initially — on the dominant human sense (vision)[69]. Also, how multiple sensory inputs relate to emotional responses is poorly understood[70]. Therefore, comparisons of different GI scenarios through vision may have greater ecological validity than multisensory comparisons until empirical evidence on the mutualistic effects of vision, sounds and scents are available[70].

2.1 Computer-Generated Environments

Different VR systems require different programs to create and display GI content. For models and animations that are purely generated by computers, the content is first generated in a game engine (i.e., Unity 3D, Autodesk 3ds Max, or Maya)or professional modeling program (i.e., SketchUp or Autodesk Revit). Game engines have built-in functions that can model, render, and generate VR environments in a single program. These engines are also particularly powerful platforms to create environments with user interaction, because they can simulate several scenarios and provide many navigation/interaction options. Early attempts at using game engines to create GI content in VR can be found in studies of environmental impact assessments. For example, Gang, Choi, Kim, and Choung[71]invented a tool kit using Unity 3D and WebGIS system to display hydrological and water hazard information in VR. The basic concepts and procedures of using a game engine (Unity 3D) to create VR content is described elsewhere[72].

The second approach to creating computergenerated environments is may be more accessible to researchers who are newer to VR. The work flow starts with a Building Information Modelling(BIM) process and converts these models to VR environments. The actual VR experience is delivered through a mobile app (i.e., InsiteVR or Kubity) or computer-based platform (i.e., eyecad VR or Enscape). The major difference between the phone- and computer-based platforms is that the former converts the entire VR model while the latter delivers real-time renderings. An example of this workflow can be found in a recent study of spatial cognition and wayfinding assisted through hospital gardens[73]. Figure 3 illustrates the quality of real-time renderings of the hospital gardens as viewed by the VR user.

Researchers can blend the two workflows(game engines and BIM) to create and deliver GI content in VR as well. Yu, Behm, Bill, and Kang[74]used Autodesk 3ds Max and Unity 3D to study differences in visual and noise impacts between ambient wind park soundscapes. Similarly,Lin, Homma & Iki[75]generated 3D models with ArcGIS 10.0 and Adobe Photoshop CS6 to examine people's visual preferences for different sizes of blue spaces.

2.2 Computer-Modified Real Environments

An entirely different approach to creating VR content is recording pre-existing environments and digitally editing them. The real space is captured with a special camera with two or more fish-eye lenses capable of capturing 180º videos. The videos are later stitched together and synchronized. They can be edited to add, remove, or edit GI elements(i.e., trees, grass, buildings, or park benches) using video or photo editing software programs, such as Adobe Photoshop, Premiere, or After Effects. The editing work flow may be familiar to researchers of the built environment who have edited 2D images or videos[44]. Multiple videos of GI options can also be taken and only minimally processed. For instance,Hedblom et al.[68]removed the tripod, sharpened the image, and corrected the light in Adobe Photoshop. Calogiuri et al.[54]used Adobe After Effects, Warp Stabilizer VFX, and Samsung Gear 360 ActionDirector to improve the video stability.Also, Browning et al.[52]reported no edits to their videos other than minor lighting changes. Figure 4 provides examples of how 360º videos can be edited to improve the imagery and modify GI elements.

3 VR Study Design

If researchers want to state with confidence that a certain GI option improves human health,they may be best served with an experimental research design. Experiments differ from other types of designs, such as correlational or quasiexperimental studies, in the use of randomlyassigned interventions that are controlled by the researcher. Without randomization and a control group, it is difficult to claim that an exposure to a specific condition caused observed changes in participants. That said, many types of studies can and are used to evaluate the impacts of design on health using VR[76]. Here, we discuss experimental studies since they are most appropriate for comparing multiple environments in VR and can show cause-and-effect relationships.

Experimental research falls within two categories: between-subjects and within-subjects designs. Both are suitable options for investigating the benefits of GI exposure. We cover these categories below as well as threats to the validity of the collected data and how these might be addressed.

3.1 Between-Subjects Designs

Between-subjects designs involve random assignment of each participant to a group.Participants in each group are treated identically except for the variable of interest. One group receives a treatment or condition (i.e., a VR experience with GI present) while the other group receives an alternative condition (i.e., an experience without GI present or with a different form of GI).Randomization is a necessary step that minimizes the chances that participants in each group do not differ in any systematic way. Commonly, simple randomization is performed, and each participant has equal odds of being in group A or B (or C and so on). However, when sample sizes are small,block randomization can be used to ensure equal numbers of participants in each group based on confounding factors[77]. Psychology and biomedical research have focused on equal number of male and female participants, but other confounding factors may also be important to balance in GI studies using VR. These include but are not limited to age, race/ethnicity, socioeconomic status,familiarity/preferences for certain landscapes,time spent in natural settings, connection with nature, perceived safety, VR experience, and VR presence[46].

3.2 Within-Subjects Designs

Within-subjects designs (also called repeated measures designs) involve each participant undergoing multiple experimental conditions.Participants effectively serve as their own control,which increases statistical power and ability to detect weak effects. (Effects describe the size of differences between experimental conditions,such as exposure to different VR environments.)Such advantages make within-subjects designs more desirable for some experiments in which participants are exposed to simulations of natural settings, including GI[46].

There are downsides of within-subjects designs. Often, the involves multiple visits to a laboratory, and repeated visits can lead to participant attrition, which reduces sample size.Also, study results are more susceptible to bias from order and carryover effects. These describe how participants learn to better perform on cognitive tasks with repeated exposure, as well as how participants become fatigued over the course of an experiment and provide less reliable answers to questionnaires as the experiment progresses.

These biases can be partially overcome by counterbalancing the order in which experimental conditions are administered. For instance, half of participants could be exposed to scene A and then to scene B, and the other half are exposed to the scenes in the reverse order. Allowing participants to practice questionnaires or cognitive tasks ahead before the experimental starts may also prevent carryover effects[78].

3.3 Threats to Validity

Experimental study designs are subject to several types of bias. Demand characteristics refer to a participant's tendency to adjust their behavior and responses to questionnaires based on the perceived desires and expectations of the researcher. This type of bias is of particular concern for within-subject experiments during which participants experience multiple experimental conditions and are likely to identify the differences between each[79]. Researchers also have a tendency to adjust their behavior based on their own desires and expectations. Such a bias is called an expectancy effect, and it is reflected in experimenters expressing or withholding enthusiasm for one condition over another in their speech and body language[80].

These biases can be reduced by blinding the participant and/or the researcher to the condition.Blinding the participant involves withholding the condition(s)/treatment(s) that other participants are receiving. Blinding the researchers involves withholding the condition/treatment that each participant is assigned. Double blinding both the participant and researcher is most effective at reducing these biases but difficult or impossible in VR studies. One approach to double-blinding the condition is by using a within-subjects design and programming the HMD to present virtual environments in a randomized order that is different for each participant. The software must also record the order of presented environments for each participant.

4 Health Outcome Measurement

Numerous health outcomes may be expected to occur from exposure to GI in VR. For example,green walls and views of extensive forest cover in VR have been associated with stress recovery and anxiety reduction[11]. Eucalyptus trees, meadows,and streams have improved mood states better than an urban setting without GI[81]. Dense forests and meadows have also improved mood states or affective arousal compared with an indoor settings[52], abstract paintings[82], or urban centers without GI[83]. Urban parks and forests have improved mood and attentional states better than barren landscapes with buildings and trees only in the background[84]. The stress-reducing potential of street trees have been documented in multiple studies[38-39]. Across these articles, the findings have been measured with both objective measures(physiological responses) and subjective measures(self-reported responses). Both types of measures can be used to examine either physical health or mental health/cognitive performance outcomes.

4.1 Physical Health Outcome Measurement

Physical reactions to environmental stimuli are primarily measured with devices that measure physiological changes. These changes involve either cardiovascular responses, including blood pressure,blood volume pulse (BVP), heart rate, and heart rate variability (HRV)[85], or hormones (i.e., cortisol in the saliva or blood).

Both cardiovascular responses and hormones yield attractive and usable data that link to longterm physical health. Elevated heart rates for long periods of time increase the risks of blood clot, high blood pressure, stroke, and other cardiovascular diseases[86]. Also, these indicators serve as objective measurement of stress[87], and chronic stress can lead to several physical and mental health issues such as insomnia, reproductive problems, or cancer[88].

Physiological data give accurate results if measured corrected and are relatively easy to collect with increasingly low-cost wearables[85]. However,the resulting data can be difficult to analyze,particularly if the sensors do not record data throughout an experiment[89]. In regard to hormone biomarkers, data collection can also be intrusive and stressful to the participants and requires extended experimental timing. Salivary cortisol, for example, requires 20 minutes to travel from blood to the mouth and participants will be required to remain in the laboratory setting throughout this rest period[90]. Time of day, food, tobacco and caffeine use, physical activity, sleep, and gender may also affect fluctuations in cortisol levels and other biomarkers[91-92]. Accurate data collection of these measures requires particularly careful and rigorous protocols to control for myriad confounding effects.

4.2 Mental Health and Cognitive Function Outcome Measurement

Changes in mood and attention/working memory are commonly studied outcomes of GI exposure[32,46]. These outcomes can be measured with physiological measures as well as standardized psychological survey batteries. Cognitive function is considered a form of mental health that includes the ability to process information, make decisions,and succeed in life[93].

Physiological measures include bodily responses to stress and negative moods. These include facial movements, such as frowning with forehead muscle tension[94], sweating through skin conductivity[95-96], and activity in the brain[97].The advantage of these measures is that they are involuntary and not influenced by participants.They can measure moods with objectivity, but some are confounded by other bodily processes, such as cognitive load in the case of neural activity[98]. Like physiological measures of physical health, these data can be difficult to analyze and, in the case of brain activity, expensive to collect.

Many standardized survey approaches have been psychometrically validated and are readily available to researchers. Surveys are typically deployed both before and after exposure to each VR environment. Change scores are calculated to determine whether the environment influenced the outcome of interest. Common survey approaches related to emotional health are summarized in Tab. 1.

Attention and working memory can be measured with tasks that are performed after VR exposure but are commonly accompanied with a pre-exposure depletion exercise. These exercises control the antecedent condition from which the participant's cognitive abilities can be measured[99]in accordance with Attention Restoration Theory,which is a common theoretical explanation as to why green infrastructure is beneficial to health[13].Pre-condition tasks can also be applied for other outcomes, such as perceived stress[100-101],in accordance Stress Restoration Theory. This provides an alternative explanation of why GI benefits human health related to evolutionary psychology[102].

Additional technologies are available but underutilized in research on the health effects of GI. Notably, built-in eye-tracking can categorize GI features by participant response[11]. An example of a device with built-in eye-tracking is the tethered HTC Vive Pro Eye. This device records foveal attention throughout each 360-degree or 3D environment using multiple parameters(saccades and fixations). Data can be measured simultaneously with biosensor data (i.e., skin conductivity) to objectively determine to what extent different GI features demand participant's attention, are preferred by participants or activate emotional responses.

5 Safety Considerations

Conducting VR research generally poses little risk to subjects. However, there is always the possibility of cybersickness. Also, as a result of the recent COVID-19 pandemic, communicable disease transmission has emerged as a safety concern.

5.1 Cybersickness

Cybersickness describes a variety of possible symptoms similar to motion sickness that can be caused either by vestibular stimulation (physical movement) or visual stimulation (observed movement) in VR[116]. Symptoms may include eye strain, headache, pallor, sweating, dryness of the mouth, fullness of the stomach, disorientation,vertigo (disordered state characterized by surroundings appearing to swirl dizzily), ataxia(lack of coordination), nausea, vomiting, dizziness,salivation, and burping[116-117].

Numerous determinants of cybersickness have been long discussed in the scientific literature,such as user characteristics, movement in the VR environment, duration and frequency of exposure,and device specifications (i.e., frame rate, field-ofview, and resolution)[118]. Also, the extent to which the VR user feels like they are “being there” in the virtual space (presence) is negatively related to cybersickness symptoms[119]. Based on these determinants of cybersickness, proposed solutions include use of high-quality, comfortable HMDs and minimizing the number of visual complexity of objects in the three-dimensional (3D) environments(i.e., the number of triangles, vertices, and texture sizes) as much as possible without sacrificing realism to reduce latency and optimize frame rate[76,120].

5.2 Communicable Disease Transmission

VR research often involves placing a device with porous materials near people's eyes, nose and mouth. This operation can quickly spread bacteria, viruses, and fungi between participants.Appropriate cleaning and pre-screening of participants is a necessary process of VR research,and inadequate adherence to these safety protocols may cause human subject review boards to terminate a project, or worse, HMDs may serve as a vector of disease.

The foam inserts that sit inside the HMD soak up bodily fluids, such as sweat. Disposable face cover masks can be used and switched between users. Replacement foam inserts can be bought for approximately the same amount (10 to 20 USD per insert).

Hard surfaces of the HMD should also be cleaned after every use②. For greatest convenience,boxes may be used with specially placed ultraviolet(UV) lights that are reported to kill 99.99% of bacteria, viruses, and fungi from every surface of the headset in 60 seconds③.

6 Conclusion

Virtual reality provides an exciting,convenient, and relatively inexpensive method for researchers to test the human health benefits of green infrastructure prior to implementation.Continued use of VR research to study health promotion through urban planning and design,recreation, engineering, and other aspects of the built environment may help reduce the global burden of physical and health conditions as the globe becomes increasingly urbanized.

Notes:

① https://www.vrs.org.uk.

② For more information on the safe use of HMDs, we recommend visiting postings from The Academy of the International Extended Reality (https://aixr.org/press/articles/covid-19-safety-for-virtual-augmented-realityaixr-guidelines/) and a summary of the Facebook's recommendations of Oculus hardware (https://uploadvr.com/sanitize-clean-vr-headsets-oculus/).

③ https://www.cleanboxtech.com.

Acknowledgments:

We would like to thank Rutali Joshi, Ph.D. candidate and Uniza Rahman, graduate student in Architecture + Health at Clemson University for developing Figure 4 in the paper.

Sources of Figures and Table:

Fig. 1 © Pongsakorn Suppakittpaisarn; Fig. 2-3 © JIANG Shan; Fig. 4 © Rutali Joshi and Uniza Rahman; Tab. 1 ©Matthew Browning.

(Editor / WANG Yilan)