[Paper Review] Applications of Robots for COVID-19 Response

Robin R. Murphy, Vignesh B.M. Gandudi+1 more|arXiv (Cornell University)|Aug 16, 2020

Disaster Response and Management1 references40 citations

TL;DR

This paper catalogs 262 reports of COVID-19–related robot use, identifying 203 concrete instances across 104 robot models in six application categories, highlighting teleoperation dominance and gaps in autonomy.

ABSTRACT

This paper reviews 262 reports appearing between March 27 and July 4, 2020, in the press, social media, and scientific literature describing 203 instances of actual use of 104 different models of ground and aerial robots for the COVID19 response. The reports are organized by stakeholders and work domain into a novel taxonomy of six application categories, reflecting major differences in work envelope, adoption strategy, and human-robot interaction constraints. Each application category is further divided into a total of 30 subcategories based on differences in mission. The largest number of reported instances were for public safety (74 out of 203) and clinical care (46), though robots for quality of life (27), continuity of work and education (22), laboratory and supply chain automation (21), and non-clinical care (13) were notable. Ground robots were used more frequently (119) than aerial systems (84), but unlike ground robots, aerial applications appeared to take advantage of existing general purpose platforms that were used for multiple applications and missions. Of the 104 models of robots, 82 were determined to be commercially available or already existed as a prototype, 11 were modifications to existing robots, 11 were built from scratch. Teleoperation dominated the control style (105 instances), with the majority of those applications intentionally providing remote presence and thus not amenable to full autonomy in the future. Automation accounted for 74 instances and taskable agency forms of autonomy, 24. The data suggests areas for further research in autonomy, human-robot interaction, and adaptability.

Motivation & Objective

• Document the actual uses of robots during the early COVID-19 period (Mar–Jul 2020) to guide policy makers and roboticists.
• Develop a taxonomy of robot applications by stakeholders and work domain.
• Analyze modality (ground vs aerial), control schemes (teleoperation, automation, taskable autonomy), and autonomy levels.
• Identify gaps and priorities for future research in autonomy, HRI, and adaptability.

Proposed method

• Manual content analysis of 262 reports from press, social media, and scientific literature.
• Extraction of 203 explicit COVID-19–related robot usages across 33 countries.
• Classification into a six-category taxonomy with 30 subcategories based on work domain and mission.
• Aggregation by modality (ground vs aerial) and by control scheme (teleoperation, automation, taskable autonomy).
• Identification of frequently used models and commercial availability status of robots.

Experimental results

Research questions

• RQ1 What categories of COVID-19–related robot applications emerged across stakeholders and work domains?
• RQ2 How do ground and aerial robots differ in usage, deployment, and platform maturity during the pandemic?
• RQ3 What control schemes (teleoperation, automation, or semi-autonomous taskable agency) dominated COVID-19 robot deployments?
• RQ4 What gaps exist in autonomy and human-robot interaction that warrant future research?
• RQ5 Which robot models were most commonly used and to what extent were they commercially available or modified?

Key findings

• There were 203 concrete robot deployments across six application categories.
• Ground robots (119 instances) were more common than aerial robots (84 instances).
• Public Safety (74 instances) and Clinical Care (46 instances) were the largest categories; others included Quality of Life, Continuity of Work and Education, Laboratory and Supply Chain Automation, and Non-Hospital Care.
• Teleoperation dominated control (105 instances), with many cases emphasizing remote presence rather than full autonomy.
• Automation accounted for 74 instances, with 24 involving taskable autonomous forms of autonomy; several subcategories featured both autonomous and teleoperation elements.
• A wide variety of models and platforms were used (104 models; 82 commercially available or existing prototypes; 11 modified; 11 built from scratch), with DJI Mavic 2 Enterprise, TEMI, and MMC Among the most common models.
• Aerial platforms were often general-purpose, enabling quick repurposing across missions; ground robots tended to be more mission-specific (e.g., disinfection, dispensing).
• The data highlights opportunities for advances in navigation in large outdoor and cluttered indoor spaces, manipulation capabilities, physical human-robot interaction, and robust computer vision for inventory and recognition.