An Analysis of International Use of Robots for COVID-19
Robin R. Murphy, Vignesh B.M. Gandudi, Trisha Amin, Angela Clendenin, Jason Moats
AAn Analysis of International Use of Robots forCOVID-19
Robin R. Murphy
Depart. of Computer Science & EngineeringTexas A&M UniversityCollege Station, TX 77843-3112 [email protected]
Vignesh B.M. Gandudi
Depart. of Computer Science & EngineeringTexas A&M UniversityCollege Station, TX 77843-3112 [email protected]
Trisha Amin
School of Public HealthTexas A&M UniversityCollege Station, TX 77843-3112 [email protected]
Angela Clendenin
Department of EpidemiologyBiostatisticsTexas A&M UniversityCollege Station, TX 77843-3112 [email protected]
Jason Moats
Texas A&M Engineering Extension ServiceCollege Station, TX 77843-3112 [email protected]
Abstract
This article analyses data collected on 338 instances of robots used explicitly in response toCOVID-19 from 24 Jan, 2020, to 23 Jan, 2021, in 48 countries. The analysis was guidedby four overarching questions: 1) What were robots used for in the COVID-19 response?2) When were they used? 3) How did different countries innovate? and 4) Did havinga national policy on robotics influence a country’s innovation and insertion of robotics forCOVID-19? The findings indicate that robots were used for six different sociotechnical workdomains and 29 discrete use cases. When robots were used varied greatly on the country;although many countries did report an increase at the beginning of their first surge. Tounderstand the findings of how innovation occurred, the data was examined through thelens of the technology’s maturity according to NASA’s Technical Readiness Assessmentmetrics. Through this lens, findings note that existing robots were used for more than 78%of the instances; slightly modified robots made up 10%; and truly novel robots or novel usecases constituted 12% of the instances. The findings clearly indicate that countries witha national robotics initiative were more likely to use robotics more often and for broaderpurposes. Finally, the dataset and analysis produces a broad set of implications that warrantfurther study and investigation. The results from this analysis are expected to be of valueto the robotics and robotics policy community in preparing robots for rapid insertion intofuture disasters. a r X i v : . [ c s . R O ] F e b Introduction
The COVID-19 pandemic provides an unique opportunity to examine the in situ operationalization of robotsduring a pandemic, and, by extension, disasters in general. Several articles have already been published thatattempt to summarize how robots have been widely used to mitigate the medical, economic, and socialimpacts of COVID-19, notably (Murphy et al., 2020; Murphy, 2020; Shen et al., 2020). More often, surveyarticles such as (Clipper, 2020; Madurai Elavarasan and Pugazhendhi, 2020; Mardani et al., 2020; Yanget al., 2020) focus on how robots could be used and on technological gaps or deficiencies preventing suchuses.While such analyses are valuable to the robotics community, they neglect larger questions as to whetherthere are patterns of how different countries have employed robots to cope with COVID-19. Establishingwhich robots have been used for what purposes and how quickly they were deployed provides a baseline forfuture work. Analyzing the influences on why and how those robots were put into practice is valuable indetermining research investments and national policies to prepare for effective response to future disasters.In order to provide this larger view of the use of robotics for COVID-19, this article surveys the state of thepractice and addresses four specific questions: • What were robots used for?
Investigating what robots were used for during the coronavirus pandemicprovides a de facto work domain analysis that can guide R&D for future pandemics and possibly fu-ture disasters in general. Variations in applications by country may help identify cultural, economic,and policy differences that impact adoption and economic markets. • When were robots used?
Documenting whether robots were operational before, contemporaneouslywith, or lagged after the surge for a country captures the overall readiness of robots for a disaster. • How did different countries innovate?
Documenting the types of robotics innovations exhibitedby individual countries to cope with pandemics can inform research, development, and technologytransfer for future disasters. This understanding can be used to craft appropriate policies, includingnational robotics initiatives. • Did having a national policy on robotics influence a country’s innovation and implementation ofrobots for COVID-19?
Six countries have national or economic union initiatives in robotics. Whilethose policies differed in scope and size, the existence of a formal commitment to robotics could beexpected to foster rapid insertion or adaption for a disaster. Documenting the performance of thosecountries with initiatives can be useful for creating new or modifying existing policies.The analysis in this article is based on the open source Robotics For Infectious Diseases (R4ID) dataset(Robotics for Infectious Diseases, 2021). R4ID captures the use of ground, aerial, and marine robots fromthe first reported instance, 24 Jan 2020, to 23 Jan 2021, covering a calendar year. As of 23 Jan 2021, R4IDcontained 338 distinct instances of robots being used explicitly to cope with the COVID-19 pandemic. Theinstances were from 48 countries in six continents, Africa, Asia, Australia, Europe, North America, andSouth America. Table 1 provides a list of countries, while Figure 1 shows those 48 countries on a map.In order to examine international trends in more detail, this article provides summative data for all 48countries and delves deeper into the eight countries with the largest number of instances of robot use: US,China, India, Great Britain, Italy, South Korea, Spain, and Singapore. Thailand is tied with Singapore fornumber of instances, but for the purposes of this article and space limitations, Singapore is considered asoccupying the eighth position because it has an earlier date for the first use. Two of these eight countries,Italy and Great Britain are also among the ten countries with the highest per capita incidence of COVID-19 deaths (Worldometers, 2021). Thus, the top eight provides a tractable number for investigation whilerepresenting countries with large COVID-19 impacts and sizable deployments of robots. Furthermore, sixable 1: List of countries organized by number of instances of robot use for COVID-19 with the associatedfirst date of use. * indicates that the date was inferred.of the eight countries, China, Great Britain, Italy, Spain, South Korea, and the US, have formal roboticsinitiative programs, possibly providing insights into the impact of national policies.The article is organized as follows. It begins with a description of the R4ID dataset and the data collectionand analysis methodology. Next, each of the following sections explores one of the motivating questionsby presenting relevant data and then extracting findings. The article concludes with observations for therobotics and policy communities.
The robot instances analyzed in this article are extracted from the R4ID database. Each instance documentsan operational robot explicitly used for COVID-19, the robot’s modality, manufacturer and model, geographiclocation of use, and date of use. The database is restricted to operational robots, which are defined as thosesituated in the work place, not a laboratory or simulation, and in use by end-users, not developers, for actualuse, not a strictly controlled experiment. This eliminates many of the speculative robots currently underdevelopment cited in the surveys by (Clipper, 2020; Madurai Elavarasan and Pugazhendhi, 2020; Mardaniet al., 2020; Shen et al., 2020; Yang et al., 2020) and allows the analysis in this article to be based on actualdocumented use.The R4ID instances enable the analysis of when robots were used and cumulative frequency plots by country.The extracted instances were clustered into sociotechnical work domains and use cases using the constantcomparative method of quantitative analysis (Glaser, 1965; Glaser and Strauss, 1967; Merriam, 1998; Ruona,
Figure 1: 48 countries reporting use of robots for COVID-19 shown in dark blue; map generated throughMicrosoft Excel.2005) to describe what robots were used for internationally. Each robot instance were also rated for technicalmaturity using the NASA Technical Readiness Assessment process (Hirshorn and Jefferies, 2016; Frerkingand Beauchamp, 2016) that will be described in more detailed in Section 5.The roboticsForInfectiousDiseases database was created by weekly searches of online press reports, socialmedia, and the scientific robotics and medical literature. The searches began in March, 2020, when apandemic was declared. English phrases and keywords (COVID, COVID19, COVID-19 robots, COVID19Robots, COVID 19 Drone, COVID 19 UAS, COVID Drone, COVID UAS, “COVID-19 and Robots”, “Useof Robots for COVID-19”, “Use of Robots for the present pandemic”, and “COVID-19 Robot uses”) wereused. The search often uncovered non-English reports as “robot” and “COVID” are generally expressed asthose words regardless of language. In addition, the comments section of the social media posts, and pressreports were manually scraped as well to obtain additional links. Every effort was made to determine thedate of actual use of a robot, and if that was unknown, it was noted and the date was considered the dayof the report for analysis purposes. The search discovered the first reported instance of a robot being usedfor COVID-19 in the world occurring on 24 Jan, 2020, in a hospital in Seattle, Washington, US (A mandiagnosed with Wuhan coronavirus near Seattle is being treated largely by a robot, 2020). The searcheswere discontinued on 23 Jan, 2021, providing a calendar year of robotics reports.The database contains 424 reports leading to 338 distinct instances of use. The 424 reports were filteredto retain only the 400 reports that discussed actual robots put into operational use explicitly because ofCOVID-19. The reports often did not given any indication of the length of operational use or quality of use;therefore, the instances include situations where the robot was probably a working demonstration ratherthan placed in regular service. However, this filtering did eliminate reports about robots in laboratoriesor under development with no explicit date of deployment. The 400 reports contained duplicates, eitherretweets and repostings or a different news agency covering the same robot. These were merged, resulting in272 entries. Since several news articles described multiple robots or how a model of robot had been used formultiple applications, those entries were split into 338 separate instances of a robot and application tuple.While every effort has been made to produce a comprehensive, replicable search, there are numerous limita-tions of the R4ID dataset. For example, the R4ID dataset is most certainly incomplete and noisy. Robotsn service may not have been the subject of press or research reports and thus skipped. Also, the use ofEnglish keywords may have likely limited discovery, though as noted earlier “robot” and “covid” appear inthose forms in many non-English postings. From both a robotics and epidemiological standpoint, the datais usually coarse. The robotics reports typically did not provide technical descriptions of the work envelope,though the general characteristics could be inferred from photos and videos, metrics used to evaluate theperformance of the robots, or the duration of the operation. However, the volume of reports suggest thatit is sufficient to detect general trends in robotics and robotics policy and to identify topics for further in-vestigation. It is difficult to precisely correlate lockdown and surge dates with robot instances as surges andlockdowns were often local rather than national, multiple media sources can be inaccurate, and lockdownsdid not occur in some countries. However, examining the trends in cases and deaths for entire countriesprovides helpful insights.The 338 instances were iteratively clustered into sociotechnical work domains and use cases using the con-stant comparative method of quantitative analysis (Glaser, 1965; Glaser and Strauss, 1967). The groupinginto sociotechnical work domains was based on similarities in the stakeholders and regulations impactingimplementation, the end-user skills and expectations, the general work envelope, and the general mission orobjective for having robots.. Six sociotechnical work domains emerged: • Clinical Care . This refers to hospital and patient related functions that generally occur in struc-tured medical environments with specialized personnel. These activities are regulated by the govern-ment and medical insurers who ultimately must approve the use of robots and cost reimbursement.The term
Clinical Care is used in its broadest sense and includes intensive care units as well aspatient assessment and the diagnosis and treatment processes discussed in robot surveys for medicinesuch as (DiLallo et al., 2021). • Non-hospital Care.
This covers medical facilities outside of hospital care, such as quarantinecamps, nursing homes, and private clinics. These facilities share some of the trained personnel andprocesses as
Clinical Care but operate under different budgets and regulations. • Laboratory and Supply Chain Automation.
This sociotechnical work domain represents themedical support industry which performs specialized functions such as processing tests for coro-navirus and supplying hospitals and public health workers. While they are part of the medicalresponse, they are economically and functionally distinct. • Public Safety . This refers to public health activities carried out by law enforcement or otheremergency management agencies. These activities include disinfecting public spaces and enforcingquarantine restrictions by trained personnel. • Continuity of Work and Education . This sociotechnical work domain encompasses activitiesby private businesses and educational institutions to maintain operations. While businesses andeducational institutions have different funding streams, they share similar work envelopes and haveto contend with workers or students who may not be trained to use or interact with robots. • Quality of Life.
This sociotechnical work domain refers to the ways in which robots are usedto support individuals, either by private companies delivering goods and services or facilitatingpandemic-appropriate social interactions. Individuals themselves may use personal robots in newways to cope with pandemic lockdowns, for example, walking a dog with a small unmanned aerialsystem.The constant comparative method was also used to cluster the instances within a sociotechnical work domaininto use cases. These use cases are shown in Figure 2 with largely self-explanatory labels. Perhaps themost confusing labels are for
Observational Telepresence and
Interventional Telepresence in Clinical Care ; observational refers to non-contact interactions such as remote patient assessment whilenterventional refers to teleoperating a robot that makes physical contact with a patient. It should benoted that several work domains have a delivery use case. These are considered different use cases formany reasons. Consider robot delivery of medical supplies in a hospital ward (
Clinical Care ) differs froma small aerial vehicle dropping of a lightweight package of test vials at a laboratory (
Laboratory andSupply Chain Automation. ), which is in turn different from a robot car dropping off a week’s worth ofgroceries to a surburban household (
Quality of Life ). Likewise disinfecting outdoor public spaces or largeindoor spaces such as covered stadiums (
Public Safety ) poses different technical challenges and personnelthan disinfecting a hospital room (
Clinical Care ) or sanitizing a office (
Continuity of Work andEducation ). Public SafetyDisinfecting public spacesQuarantine EnforcementIdentification of Infected
Public Service Announcements
Monitoring traffic flow D e c r ea s i ng f r equen cy o f r epo r t ed u s e s Clinical CareDisinfecting Point of CareObservational TelepresenceDelivery & InventoryPatient & family socializingInterventional TelepresencePatient intake & visitors Non-Hospital CareDelivery to quarantinedQuarantine SocializingPublic Health Surveillance
98 85 68
Laboratory and Supply Chain AutomationDeliveryLaboratory AutomationInfectious mat. handlingManufacture or Decon PPE
41 10
23 1 41 1 4 274
Quality of LifeDelivery food
Delivery non -food purchases
Other personal activitiesInterpersonal socializingAttend public social events Private SecurityContinuity of Work and EducationSanitation work/schoolWH & Process automationTelepresencePrivate Health SurveillanceConstruction & Agriculture
Figure 2: Instances of ground (brown), aerial (blue), and marine (gold) robot by sociotechnical work domainand use case.Figure 2 summarizes the 338 instances by use of robots by sociotechnical work domain, use case, andmodality. Note that two earlier versions of this visualization (Murphy et al., 2020; Murphy, 2020) haveappeared previously. This version has more instances than and due to the content of the instances, theclustering of work domains and use cases changed. The top row shows the number of instances for that workdomain. The work domain icons are ordered by decreasing frequency, left to right. Each icon has sidebarsdecomposing the instances into aerial (117), ground (219), and marine (2) vehicle modalities. Under eachdomain icon is a column of use cases, with the icon for each use case also containing a sidebar with numberof instances by modality.Note that this criteria for comparison based on sociotechnical work domains and then by use case produces adifferent taxonomy from other surveys. The taxonomy in this article is most similar to the one by (Cardonaet al., 2020), which appears to be based on one of the two earlier iterations of the R4ID database (Murphyet al., 2020). Cardona et al. restricts their survey to operational robots but provides only nine robots, muchless than the 338 in this article. This article’s focus on sociotechnical work domains is broader that thestrictly economic partitioning used by (Madurai Elavarasan and Pugazhendhi, 2020). The employment ofthe constant clustering method resulted in distinctly different clustering of work domains and use cases fromthe categories in (Shen et al., 2020). That survey surveys 200 robots, some of which appear to be laboratoryprototypes, and does not provide any motivation for categorization. Their categorization appears to be byfunctionality and skewed toward medical applications: diagnosis and screening, disinfection, surgery andtelehealth, social and care, logistics and manufacturing, and other. In the clustering used by this article,disinfection is a function that appears as distinct use cases in multiple sociotechnical work domains. Forexample, disinfecting indoor, sparse hospital rooms which can be closed off (
Clinical Care ) poses differentable 2: Instances of robots by country for each of the six sociotechnical work domains.robot design considerations and implementation constraints than robotic disinfection of large outdoor publicspaces in
Public Safety . All three modalities (air, ground, marine) of robots were used for wide variety of applications, spanningsix work domains and 29 use cases. While front-line uses of robots for public health were the majorityoverall and in most countries, there was an almost equally large use of robots for economic and individualapplications, especially in the top eight adopters (US, China, India, Great Britain, Italy, South Korea, Spain,and Singapore).Table 2 breaks out the data in Figure 2 for each of the 48 countries distributed by work domain. The rowsare organized in descending order of total number of instances.As would be expected, the majority of instances of use were for aspects of public health, primarily for front-line Public Safety (98) and
Clinical Care (85), with smaller numbers for
Laboratory and Supply ChainAutomation (41) and
Non-Hospital Care (10). However,
Continuity of Work and Education ,which is driven by businesses and individuals, not by public health or the medical industry, was the thirdmost reported use of robots with 68 instances. While each of the 29 use cases differed in some notable way,Figure 2 suggests that the majority of the largest use cases could be categorized as robots for disinfection orsanitation to provide effective cleaning without increasing manpower or risking exposure, telepresence robotsto protect users from exposure, and robots delegated for transport and delivery to handle increased surge inemand.The majority of robots used were ground vehicles (219), though aerial vehicles (117) were a close second,and two marine vehicles, both unmanned surface vehicles, were used for safety and security applications.Figure 2 shows that ground robots were used exclusively for
Clinical Care and
Non-Hospital Care ,while aerial vehicles were used extensively for
Public Safety .Reported uses were not uniformly distributed by country. As seen in Table 2, only two countries, China andSouth Korea, had reported instances covering all six work domains. The US had the largest total number ofinstances but none for Non-Hospital Care, perhaps reflecting cultural and economic differences in eldercareor a lack of quarantine camps employed in countries such as China. 17 countries (United Arab Emirates,Kenya, France, Nigeria, Australia, Equatorial Guinea, Rwanda, Croatia, Czech Republic, Egypt, Honduras,Israel, Kuwait, Malaysia, Mexico, Tunisia, Turkey) out of the 48 reported only
Public Safety or ClinicalCare applications. Another 14 countries (South Africa, Austria, Chile, Cyprus, Denmark, Estonia, Ghana,Jordan, Lithuania, Netherlands, Norway, Poland, Russia, Sweden) did not report any use of robots
PublicSafety or Clinical Care ; this could be an artifact of the reporting process or reflect more flexibility ininnovation for non-governmental domains.These findings suggest: • Research, development, and policies should not limit or conceptualize innovation solely to “obvious”work domains, in this case to
Public Safety or Clinical Care . The volume of instances bybusinesses and individuals is almost equal to that of public health applications. While the impactof these non-health use cases cannot be assessed from the data, the descriptions suggest that robotswere helpful in reducing economic consequences and maintaining society. • Even after one year, only 9 instances were reported of interventional telepresence (e.g., requiringphysical interaction with a human such as for mouth or nose swabbing) were reported. While suchrobots are an important topic in fundamental robotics research for medicine, the near term benefitsof more mundane and general uses of robots should not be overlooked. • Unmanned aerial vehicles may be as important as ground robots for a pandemic, but there may bebarriers to developing such technologies without a motivating disaster. While reports indicated thatemergency waivers of aviation regulations were successfully invoked, the need for waivers suggeststhat the return of aviation restrictions may restrict further development and insertion into routineoperation.
Table 1 captures the first reported date of use of a robot for COVID-19 by country. The World HealthOrganization declared a pandemic on 11 March, 2020 (WHO Director-General’s opening remarks at themedia briefing on COVID-19 - 11 March 2020, 2020). The first eight rows shows countries that had alreadywitnessed operational deployment of a robots, with the US on 24 Jan, 2020, was nearly two months beforethe pandemic was declared. Note that only three of the early adopters were also the countries with thelargest deployment, shown in boldface.Figure 3 illustrates the cumulative number of reports over time from the eight countries superimposed overthe worldwide daily case count epidemic curve. However, the pandemic reached individual countries atdifferent times and with dissimilar impacts. Therefore, it is more informative to examine the first use of arobot in a country relative to its unique epidemic curve. Figure 3 shows the cumulative plot of robot useoverlaid with the epidemic curve for the eight countries with the largest reported instances of robot use.Figure 4 and Table 1 indicate that three of the eight countries, US, India, and Singapore, began deploying
Figure 3: Cumulative number of reports over time from the eight countries superimposed over the worldwidedaily case count epidemic curve.robots before their surge, that is, their use led the surge. Four countries, China, Spain, Great Britain, andSouth Korea, saw reports of initial robot use concurrent with their surges. One of the eight exhibited a lag,Italy’s first reported use was well after its surge peak.Figure 4 also suggests that early insertions of robots do not guarantee large scale use. For example, the UShad the earliest reported operational use of a robot explicitly for COVID-19 but the cumulative plot showsthat it was closer to 11 March, 2020, before reports of robots accelerated. Those reports could be misleadingas social media could have been focusing on any unfamiliar technology for coping with the rising concern,but if this is a true trend, then it suggests that early successes may not be adequately communicated toother potential users.Figure 3 also shows that the rate of operationalization varies. All but the US and China showed a rapid risein reports of robot use followed by a plateau that persisted despite a second (or third) surge. The increasinguse rate in the US could be an artifact of the data collection methodology favoring English, however, thedata also shows China with a similar curve.These findings suggest • the application of robotics was mostly reactive, either concurrent with, or lagging, the initial localsurge. This may indicate that countries do not have sufficient existing capacity or adopters do nothave confidence in robotics except as a last resort; Section 5 will discuss this in more depth. • the rate of sustained application of robots varies by country, and could reflect economic and regula-tory frameworks and existing robotics capabilities. There are numerous ways to categorize innovation; this article uses the NASA Technical Readiness Assess-ment (TRA) system and follows NASA’s formal process for categorizing innovation by TRA (Hirshorn andigure 4: Cumulative plot of robot use overlaid with the epidemic curve for the eight countries with thelargest number of instances of robots used for the pandemic.efferies, 2016; Frerking and Beauchamp, 2016). The TRA system has two advantages. One is that it has aformal decision tree for classifying technology so that the categorization should be uniform and repeatible.The second is that it ranks the robot within the human-robot interaction context of its intended use case, notsolely by the technical maturity of the robot components. Thus, TRA is well-suited for discussing innovationand inferring why certain robots were adopted.TRA is an expansion of technical readiness levels (TRL) into a broader classification that ranks both thematurity of a platform (the earlier TRL) and the usability for the work processes (Hirshorn and Jefferies, 2016;Frerking and Beauchamp, 2016). TRA divides readiness into three categories:
Heritage, Engineering , and
New . A
Heritage system is one that is an existing proven technology being applied to a similar mission andwork envelope, thus it should not lead to any surprises in usability. An
Engineering system is a modificationof an existing proven technology for a well-defined mission and work envelope. Such a system is highly likelyto work and not introduce unintended consequences of increased human cognitive workload or hidden work.A
New system either involves new hardware, software, or a new mission or notably different work envelope.It is high risk because it is unknown if it will work reliably and without introducing unanticipated demandson the user. Table 3: Technical Readiness Assessment of robots by country.Following the categorization method in (Hirshorn and Jefferies, 2016; Frerking and Beauchamp, 2016),Table 3 shows the distribution of Heritage, Engineering, and New robots by country, arranged in descendingrder of total instances of robots. The overall pattern is for countries to deploy existing robots for establisheduse cases (Heritage) for 78% of the instances, adapt or repurpose existing robots for established use cases(Engineering) for 10%, and create novel robots or address novel use cases (New) for 12% of the instances.Table 4: Comparison of technical readiness of robots by top eight countries.
Country Heritage Engineering NewInstances Percentage Instances Percentage Instances PercentageAll Countries 265 78% 34 10% 39 12%
US 81 85% 9 10% 5 5%CN 58 81% 8 11% 6 8%IN 17 52% 9 27% 7 21%GB 16 100% 0 0% 0 0%IT 9 69% 0 0% 4 31%KR 10 83% 1 8% 1 8%ES 9 75% 1 8% 2 17%SG 5 72% 1 14% 1 14%
Table 4 details the innovation for the top eight countries. While six of the eight countries generally followthe global trend, India stood out for having a larger percentage of Engineering systems, while Italy had alarger percentage of New systems. An explanation for why India and Italy would differ from the generaltrend is unclear and worth investigating.
Figure 5: Cumulative plots of robot use by technical readiness for all countries overlaid with the worldwideepidemic curve.The data shows a pattern of innovation over time. Figure 5 plots the reports of Heritage, Engineering, andNew robots by all 48 countries over time. The cumulative graph shows that Heritage robots were deployedfirst, as would be expected given that users would work with what was available and familiar. However,surprisingly, it indicates that Heritage uses continued to grow over the year and outpace the growth rateof Engineering and New. The data suggests that novel Engineering and New innovations have not beensustained. Figure 6 provides plots of innovation for the top eight adopting countries; the eight, especiallythe top four, follow the pattern in Figure 5. Returning to Table 2 and Table 1, it can be seen that smallercountries, such as Rwanda and Poland, typically add Heritage robots; this continued addition of Heritagesystems may also explain why Heritage reports continue to increase over time.The data also shows that robots were reported for multiple sociotechnical work domains from the verybeginning of the outbreak and from the beginning of a country’s use. Figure 7 illustrates that withinapproximately a month, robots were being fielded for multiple work domains and not restricted to
Clinical
Figure 6: Individual country cumulative plots of robot use by technical readiness overlaid with the epidemiccurve for the eight countries with the largest instances of robots for the pandemic.
Care . Figure 8 shows that operational robots in multiple work domains were announced within one monthof that country’s initial foray.
Clinical Care was not necessarily the first reported use of robots, eventhough a pandemic is a public health crisis. Three of the top eight countries reported Quality of Life (China)or Public Safety (Spain, Italy) as the first applicationThe findings on innovation are that: • Heritage robots, that is, existing robots for known use cases, overwhelming dominated implemen-tations, accounting for 78% of robots. Engineering and New systems accounted for only 10% and12% respectively, suggesting that even for a long lasting disaster, users adopt existing systems whiledevelopers become immediately engaged in near-term solutions or not at all. • the global pattern for implementation over time was the immediate use of Heritage robots and thecontinued addition of Heritage robots over time. Engineering and New robots experienced an earlyburst that was not sustained. • innovation did not concentrate on public health priorities; reports of implementation of robots Figure 7: Cumulative plots of robot use by work domain overlaid with the epidemic curve for all countries.began within one month for all six work domains. This is contrast to the use of robots for highpriority public health functions in
Public Safety or Clinical Care first, then shifting to economicconcerns captured by
Continuity of Work and Education or impacts on individual citizensrepresented by
Quality of Life . This scattershot pattern of applications was the same for the USand China, suggesting that innovation is ultimately independent of government controls of technologydevelopment.
Six countries or unions have formal robotics initiatives: China, the European Union, Germany (in additionto the European Union), Japan, South Korea, and the US (How nations around the world are investing inrobotics research, 2020). Of these, Japan and Germany were not in the top eight for reported instances ofrobotics (recall they are US, China, India, Great Britain, Italy, South Korea, Spain, and Singapore. As partof the EU, Great Britain, Italy, and Spain can be considered as having a robotics initiative during this timeperiod. Therefore, of the top eight adopters, six countries had robotics initiatives and two (India, Singapore)did not.Having a robotics initiative did appear to impart some advantages. As seen in Table 2, the US, China, GreatBritain, Italy, South Korea, and Spain reported instances for more work domains than other countries, asthe average number of domains covered by a country was 2. Tables 3 and 4 indicates that at least five,the US, China, Great Britain, South Korea, and Spain, had a high percentage of Heritage robots to drawupon. This is presumably an outcome of investment in robotics and general societal awareness of robotcapabilities. However, having a robotics initiative did not ensure that a country visibly deployed robots ina timely manner. Figure 4 shows that Italy lagged behind its surge.The findings suggest: • that a national robotics initiative appears helpful in terms of prior availability of existing robots andenabling a breadth of applications. In addition, it may be helpful in creating end-user awareness Figure 8: Cumulative plots of robot use by work domain overlaid with the epidemic curve for the eightcountries with the largest instances of robots for the pandemic.and acceptance of robots. • The data is less clear as to whether national initiatives impart a clear advantage on rapid innovation,as there were relatively little need for novel robots to meet previously unknown use cases.
The R4ID dataset show that robots were used for the first year of the COVID-19 outbreak for an unexpectedlybroad set of six work domains and 29 use cases by 48 countries, many of whom are not an advanced economyaccording the International Monetary Fund (Developed Country, IMF Advanced Economies, 2021). Thissuggests that robotics is becoming more mainstream and could auger accelerated adoption by businesses andindividuals. Creating or expanding existing national robotics initiatives could increase robotics preparednessfor future pandemics or disasters. It should be emphasized that the R4ID dataset is imperfect and thus anyconclusions are speculative and offered for discussion.Returning to the questions posed in the Introduction, 219 ground, 117 aerial, and 2 marine robots, for a total38, have been documented in use in 48 countries in Africa, Asia, Australia, Europe, North America, andSouth America. Robot use was spread across six distinct work domains:
Public Safety, Clinical Care,Continuity of Work and Education, Laboratory and Supply Chain Automation, Quality ofLife, and
Non-Hospital Care . The majority of robots globally have been used in the
Public Safety (98 instances or 29% of the total) and
Clinical Care (85 or 25%) work domains, which are driven bygovernment and public health policy. protect front-line workers (including administrative staff), help copewith surge in demand for medical services However, the
Continuity of Work and Education workdomain, at 68 or 20%, was almost as large as
Clinical Care . This suggests that private industry andindividuals are acquiring and implementing robots. The majority of use cases utilized robots to protectworkers, maintain output, or to replace sick workers or enable social distancing. While the majority did notdisplace workers and would not be a threat to jobs long term, there is a possibility that increased roboticsfor warehouse and production automation will permanently displace workers.In terms of when robots were used, they were used before, during, and throughout the first year of thepandemic. Reports show that robots were in service to explicitly cope with the public health crisis incountries such as the US, India, and Singapore before the declaration of the pandemic or their local surges.It appears that the majority of countries, such as China, Spain, Great Britain, and South Korea, deployedrobots concurrent with sharp increases in their local surges, while others such as Italy lagged behind theirlocal surge. There is no clear indicator of why some countries were early adopters and others appeared todeploy robots rather late.Robotics innovation for the COVID-19 pandemic was primarily a priori innovation. Technically matureHeritage robots which had been proven in established user cases comprised the largest number of instances,265 or 78% of the total. Innovation during the pandemic appeared to follow a “low hanging fruit” strategy,where existing (Heritage) and easy to modify robots (Engineering) are used for established use cases. Newinnovation for novel use cases, such as mouth or nose swabbing, or to offer novel robot designs appears toemerge quickly but then slows over time, possibly because of the time it takes to build and program reliablerobots with high usability for new applications. It is not possible to know whether New robots would havemade a difference, but the high propensity of established use cases suggests that a pandemic does not callfor new solutions so much as rapidly scaling the availability of existing robots and end-user’s familiarity andtrust of such robots.The existence of a national robotics policy appeared to roughly correlate with a country’s use of robotsfor COVID-19. Six of the top eight countries in terms of reported instances, US, China, Great Britain,Italy, South Korea, and Spain, had national robotics initiatives. A national robotics policy appears to beassociated with earlier insertion of robots and a greater breadth of work domains that robots are appliedto. This positive association with government policy may be because many of the instances were for publichealth work domains which are funded or regulated by the government. It could also be that having anational robotics initiative is a reflection of a country’s intrinsic platform availability, talent, and awarenessof, or comfort with, robotics.It is hoped that this analysis will inform research and development of robots for the next disaster. Thedata suggests that national policies are useful and might be expanded to incentivize the development ofHeritage robots with rapid manufacturing capacity as well as platforms suitable for supporting opportunisticEngineering and New innovations.
Acknowledgments
This material is based upon work supported by the National Science Foundation under Grant IIS-2032729.Any opinions, findings, and conclusions or recommendations expressed in this material are those of theauthors and do not necessarily reflect the views of the National Science Foundation. eferences
A man diagnosed with Wuhan coronavirus near Seattle is being treated largely by a robot (2020).https://edition.cnn.com/2020/01/23/health/us-wuhan-coronavirus-doctor-interview/index.html.Cardona, M., Cortez, F., Palacios, A., and Cerros, K. (2020). Mobile robots application against covid-19pandemic. In , pages 1–5.Clipper, B. (2020). The influence of the covid-19 pandemic on technology: Adoption in health care.
NurseLeader , 18(5):500–503.Developed Country, IMF Advanced Economies (2021). https://en.wikipedia.org/wiki/Developed country.DiLallo, A., Murphy, R., Krieger, A., Zhu, J., Taylor, R. H., and Su, H. (2021). Medical robots for infectiousdiseases: Lessons and challenges from the covid-19 pandemic.
IEEE Robotics and Automation Magazine ,page to appear.Frerking, M. A. and Beauchamp, P. M. (2016). Jpl technology readiness assessment guideline. In , pages 1–10.Glaser, B. G. (1965). The constant comparative method of qualitative analysis*.
Social Problems , 12(4):436–445.Glaser, B. G. and Strauss, A. L. (1967).
The Discovery of Grounded Theory: Strategies for QualitativeResearch
Science of The Total Environment ,725:138858.Mardani, A., Saraji, M. K., Mishra, A. R., and Rani, P. (2020). A novel extended approach under hesitantfuzzy sets to design a framework for assessing the key challenges of digital health interventions adoptionduring the covid-19 outbreak.
Applied Soft Computing , 96:106613.Merriam, S. B. (1998).
Qualitative Research and Case Study Applications in Education. A Joint Publicationin the Jossey-Bass Education Series and the Jossey-Bass Higher Education Series.
San Francisco :Jossey-Bass Publishers.Murphy, R. (2020). How robots are helping with covid-19 and how they can do more in the future. In , pages 1–1.Murphy, R., Adams, J., and Gandudi, V. (2020). Robots are playing many roles in the coronavirus crisis –and offering lessons for future disasters.
The Conversation .Robotics for Infectious Diseases (2021). http://roboticsforinfectiousdiseases.org/index.html.Ruona, W. E. A. (2005).
Analyzing Qualitative Data. Chap. 14 In Research in Organizations: Foundationsand Methods of Inquiry, edited by Richard A. Swanson and Elwood F. Holton, 233-63.
San Francisco,CA: Berrett-Koehler.Shen, Y., Guo, D., Long, F., Mateos, L. A., Ding, H., Xiu, Z., Hellman, R. B., King, A., Chen, S., Zhang,C., and Tan, H. (2020). Robots under covid-19 pandemic: A comprehensive survey.
IEEE Access