ARCHIE²: An Augmented Reality Interface with Plant Detection for Future Planetary Surface Greenhouses

Abstract

More than 50 years after the last human set foot on the Moon during the Apollo 17 mission, humans aim to return to the Moon in this decade. This time, humanity plans to establish lunar habitats for a sustainable longer presence. An integrated part of these lunar habitats will be planetary surface greenhouses. These greenhouses will produce food, process air, recycle water, and improve the psychological well-being of humans. Past research has shown that a large amount of crew time, a scarce and valuable resource in spaceflight, is needed for maintenance and repairs in a planetary surface greenhouse, leaving less time for crop cultivation and science activities. In this paper, we present the concept of an augmented reality interface named ARCHIE² to reduce crew time and the workload of astronauts and remote support teams on Earth to operate a planetary surface greenhouse. ARCHIE² allows users to visualize status information on plants, technical systems, and environmental parameters in the greenhouse or other features supporting the greenhouse operations using an augmented reality headset. In particular, we report on the implementation and performance of the ARCHIE² plant detection module that runs locally on the augmented reality headset. Using images with a resolution of 320x192 pixels, arugula selvatica plants were detected using an artificial neural network (based on a YOLOv5s model) from a horizontal distance up to 50 cm with an average inference time of 602 ms and an average of 48 FPS. Based on that, the plants were augmented with labels to visualize relevant plant-specific information supporting astronauts in the maintenance of the plants.