Modern agriculture is increasingly moving towards digitization and robotics to reduce the overuse of pesticides and fertilizers while increasing yields. Key to achieving these goals is an understanding of the physical and chemical properties of the soil, which in turn leads to the need for accurate soil sampling. Traditional methods, such as manual sampling or field analysis, are uneconomical and difficult given the scale of the measurements required. Therefore, the automation of this process by means of autonomous mobile vehicles (AMVs) has become crucial.
In the framework of this master thesis, we will focus on the improvement and redesign of an existing mobile platform with a focus on the integration of a localization and navigation system. The platform will be equipped with electrodes, a device for measuring soil resistivity and a drill for sampling. As localization plays a key role in this context, we will analyse the different data processing algorithms and compare them with each other.
In the remainder of the master thesis, we discuss in detail the mobile platform, its kinematics and components. We focus on the platform management, control and implementation of sensor data fusion algorithms. We use an unscented Kalman filter (UKF) to determine the orientation of the platform, while three approaches are tested to determine the position and final orientation: an extended Kalman filter, dynamic amplification and a particle filter. We will also present the "teb local planner" plugin, the main challenges in its use and the proposed solutions. We further describe the web interface that has been designed to facilitate working with the platform. In the final chapters, we discuss the validation of the localization algorithms in a simulated environment and with real data.
|
