Soft robotic manipulation in selected agrotechnical procedures based on artificial intelligence
One of the objectives of the Specularia project is to develop a robotic manipulation system for agrotechnical interventions on plants in structured cultivation in greenhouses. The system is built upon compliant control, soft robotics, and artificial intelligence paradigms. Machine and deep learning methods are used for generation of 3D semantic and functional plant models. The generated plant models enable the system to plan robot manipulation for conducting the chosen agrotechnical interventions. The robot manipulation is planned through a compliant control algorithm based on contact force feedback. Taking into account the advantages of soft materials in applications involving manipulation of fragile objects such as plants, as well as the requirements of the selected plant manipulation activities, the adjustments of the robotic gripper and the soft tactile sensor are carried out through the development of the compliant control method, deep learning based perception methods, and soft housing.