A Comparative Ground Study of Prototype Augmented Reality Task Guidance for International Space Station Stowage Operations

摘要

Astronauts currently require extensive, near-instantaneous guidance and instruction by ground-based crew to efficiently and successfully conduct flight operations. As missions take astronauts farther away from earth and realtime communication between spacecraft and earthbound crew becomes impossible, astronauts will need technology that can help them execute flight operations with limited support. While research has shown that Augmented Reality (AR) can feasibly perform as an aid for completing certain flight tasks, there is little evidence that AR can assist in completing entire flight operations or improve flight performance metrics such as completion time. This work addresses stowage operations to investigate how AR can impact flight performance. During stowage operations, flight crew members transfer cargo items to and from different spacecraft modules. A recent stowage operation aboard the International Space Station (ISS) took 60 hours to complete with real-time ground crew support. The prolonged duration of stowage operations and the necessity for crewmembers to travel significant distances make it an appropriate domain for this investigation. StowageApp is a prototype AR application deployed on Microsoft HoloLens, and developed to assist astronauts in completing stowage operations. This paper describes the design of StowageApp and present the results of a user study comparing its performance to that of the current method of delivering stowage instructions on a handheld tablet device. This within-subject user study was performed in the ISS Node 2 Harmony, Japanese Experiment Module "Kibo," Multi-Purpose Logistics Module "Leonardo," and Columbus mockups at National Aeronautics and Space Administration (NASA) Johnson Space Center in Houston, TX, USA. Each participant completed as many of a set of predetermined stowage tasks as they could in two hours in their assigned condition. Task completion time was measured, along with the number of errors attributed to th