Verteilte autonome Systeme in der Landwirtschaft

Project team: 

 Mario Dyczka (Ph.D. Student), Prof. Dr. Nils Aschenbruck (UOS),
Dr. Sebastian Vogel (ATB), Prof. Dr. Marina Höhne (ATB)

Deploying an in-situ sensor network for monitoring crop parameters increases the temporal resolution of  measurements. These measurements can be used for calibration of remote sensors and/or agricultural growth models. On heterogeneous sites, the positions of the single sensors have a crucial impact. Leveraging existing knowledge can be used to develop sophisticated and application driven deployment strategies for the different nodes of the sensor network. The goal of this thesis is to develop optimal deployment strategies for the selection of representative sensor node positions in agricultural fields based on a-priori knowledge given by soil maps and/or remote sensing data. Moreover, mechanisms could be developed for a meaningful fusion of temporally high-resolution data from sensor networks and spatially high-resolution data from sensor mapping.

Project team: 

 Lennart Kaiser (Ph.D. Student), Prof. Dr. Nils Aschenbruck (UOS),  Dr. Thomas Hänel (UOS),Dr. Sebastian Vogel (ATB), Dr.-Ing. Volker Dworak (ATB)

Communication systems such as public land mobile networks (e.g., 4G/LTE networks) or low-power IoT networks are ubiquitous, nowadays. Temporal variations of the signal strength of such systems can be leveraged for low-cost crop and soil sensing.
The goal of this thesis is to develop new approaches for approximating crop and/or soil parameters.
These approaches will be implemented, deployed, and evaluated on agricultural fields.

Project team: 

 David Rolfes (Ph.D. Student), Prof. Dr.-Ing. Mario Porrmann (UOS), Dr.-Ing. Volker Dworak (ATB), Prof. Dr. Stefan Stiene (HOS)

Autonomous agricultural robots are a promising approach to reconciling ecological farming methods and economic yields. Modularity and collaboration are key to the success of these systems, keeping flexibility and reusability high and the overall cost low. However, developing such systems is highly challenging since it requires the co-development of mechanics, hardware, and software. This thesis focuses on the development of new methods for hardware-software integration enabling automatic configuration and resource-efficient control of modular agricultural robots.
The Robot Operating System (ROS) is a robotics middleware, which has become the de-facto standard in the context of mobile robot development in research and industry. The envisioned application scenarios require robust and efficient embedded processing solutions, e.g., for ML-based object detection and classification under hard real-time conditions with a limited energy budget. Therefore, dedicated hardware accelerators need to be integrated into the system. Reconfigurable architectures like FPGAs combine high efficiency and flexibility, making them the ideal candidates for the application domain. While new sensors and actors can be easily integrated into ROS based on the supporting software drivers, integration of hardware accelerators and automatic configuration of the combined system is an open research topic, especially concerning modular, reconfigurable robotic platforms. FPGA-based ROS nodes shall enable optimized processing for varying combinations of sensors and actuators. Additionally, runtime reconfiguration of the FPGAs shall be supported, e.g., to dynamically adapt to changing environmental conditions or to mitigate malfunctioning sensorization.

Project team: 

 Verena Tessaro (Ph.D. Student), Prof. Dr.-Ing. Mario Porrmann(UOS), Prof. Dr.-Ing. Cornelia Weltzien (ATB), Prof. Dr.-Ing. Heiko Tapken (HOS)

Joint degradation is a key factor limiting the performance of parallel kinematic robots like the Jaetrobi, developed at Leibniz Institute for Agricultural Engineering and Bioeconomy (ATB) for automatic, herbicide-free weeding using a laser. This work will focus on evaluating the long-term precision of Jaetrobi’s actuation system using a dedicated test rig with identical laser kinematics, which enables repeatable simulation of wear through defined motion patterns. Detailed analysis of joint fatigue and wear propagation will be applied toward a digital twin model that quantifies wear per movement and estimates component end-of-life. This predictive maintenance approach will support wear-optimized motion strategies, enabling synchronized joint replacement and thereby minimizing operational downtime and improving overall system reliability.