Developing a CV- and HRI-Supported Sustainable Approach for Off-Earth Robotic Planting

Abstract

The European Space Agency (ESA) funded project, Rhizome, implemented at Technical University (TU) Delft, Netherlands in collaboration with industrial partner, Vertico, focuses on the development of extraterrestrial habitats constructed from robotically printed components that are assembled into larger structures using a swarm of rovers. The habitat has an integrated greenhouse that relies among others on robotic planting. The developed Computer Vision (CV)- and Human–Robot Interaction (HRI)-supported approach implemented in collaboration with the University of Sydney (U Sydney) and the University of New South Wales (UNSW), Australia, facilitates the planting. It aims at minimizing use of human resources. Furthermore, it contributes to sustainability by (a) increasing efficiency, with robots planting seedlings faster and with greater precision than humans and by (b) improving resource utilization, with robots accurately planting seedlings at optimal spacing and depth to minimize use of water, nutrients, etc. The developed robotic approach relies on a mobile robotic system equipped with an anthropomorphic robotic hand based on soft-robotics technology, able to interact with the surrounding environment, objects, and humans. For planting, the robotic hand uses 2 Degrees of Freedom (DOF) to open and close and a camera mounted to the side of the robotic arm to capture the working area. By detecting the position and directional vector of the individual seedlings, the robotic hand is automatically positioned and suitably aligned for the seedling pickup. While a predefined automated pick-and-place code determines the planting positions, the human assists the robotic hand with gently grabbing and releasing the seedling. This approach aims to adapt robots to human interactions over time by improving their performance and responsiveness using reinforcement or imitation learning in due time. The overall goal is to advance sustainable development of habitats with integrated greenhouses through robotic technology transfer from on- to off-Earth applications and vice versa.