Optimizing Ergonomics for Robot-to-Human Object Handovers

Musculoskeletal disorders are among the most common injuries in construction workers. Thus, organisations such as the Occupational Safety and Health Administration (OSHA) in the US and the Health and Safety Executive (HSE) in the UK have invested plenty of resources into preventing work-related musculoskeletal disorders, for example, via ergonomics-aware postural risk assessment methods such as the Rapid Entire Body Assessment (REBA). However, an important gap remains in the literature for understanding how ergonomics-aware assistive robots can benefit human-robot object handovers within construction and industrial settings. In this paper, we propose a framework to optimise ergonomics for robot-tohuman object delivery and test it using a robot that is being increasingly deployed into industrial and civil scenarios: legged manipulators. First, we developed an automated vision-based ergonomics assessment technique based on the REBA assessment tool. We used this assessment to learn the relationship between ergonomics scores and object handover locations via state-of-art regression algorithms. Next, we determine the optimum handover locations and use them to optimise ergonomics for singlearm robot-to-human object delivery. We tested our framework with four different objects from a variety of construction and daily items. We compared our method to randomly generated handover locations and human-to-human demonstrations. 15 subjects performed 5 handovers for each scenario (4 objects, 3 methods), and a total of 900 handovers were examined. Our results highlight that, compared to randomly generated handover locations and human-to-human demonstrations, our framework provides better ergonomics scores and reduced muscle efforts measured by Electromyography (EMG) sensors.