Reducing human fall injuries using a mobile manipulator

Falls among elderly persons are both common and have significant negative impacts on their immediate and long-term health. In this paper, we consider whether current mobile manipulators could reduce the severity of injuries caused by human falls, and we focus on developing simple strategies toward this purpose. We present a simulation of a fall scenario that includes a humanoid and a mobile manipulator that can manipulate the humanoid at one or more points. Using metrics, we estimate the severity of injury to a human from a fall, and then we evaluate these metrics across four scenarios for interactions between the robot and the humanoid. We conduct experiments with a PR2 robot and a 20 degree-of-freedom humanoid. We parameterize the humanoid with human biometric data including mass distributions, link lengths, joint ranges of motion, and link centers of mass to approximate human fall behavior. Utilizing the full, but limited, force of its arms, the robot was able to reduce the median injury to the human as assessed by the metrics. However, scenarios where the robot´s intervention exacerbated injuries did occur, which indicates that algorithms to reduce fall damage must be carefully designed and applied to avoid worsening fall injuries.