Automation in agriculture, often referred to as Agriculture 4.0, has attracted significant interest in recent decades. The integration of technologies like artificial intelligence, robotic manipulation, and autonomous navigation, commonly used in industrial automation, also shows great potential in the agricultural sector. While these technologies are being extensively applied to automate tasks such as fruit harvesting, there is a notable lack of research focused on their application to other essential tasks, such as seeding. In this context, this article introduces an autonomous bi-manipulator robot designed to automate the seeding process in a hydroponic greenhouse. The dual-arm robotic system is capable of perceiving the environment to accurately identify the location of growing cubes within planters, securely position the cubes without causing collisions within the hydroponic system, collect seeds, and deposit them directly into the small holes of the growing cubes. Several perception and manipulation strategies are proposed and experimentally evaluated to enhance the overall process. The results demonstrate that automating this task with a dual-arm mobile robot is both feasible and efficient, achieving an average seeding time of 75 s per rock-wool cube with the most efficient strategy, including cube positioning and verification of proper seed placement.
