Publication
Spatio-Temporal-Semantic Sensor Data Management With Use Cases in Agricultural Robotics
Mark Niemeyer
PhD-Thesis, Universität Osnabrück, 2026.
Abstract
Sensor data-based approaches for autonomous machines require efficient handling of this very sensor data. Currently, no streamlined process is available (1) for capturing and storing the sensor data; (2) for the following enhancement of the data; and (3) for finally accessing specific subsets of the huge amount of stored data based on spatial, temporal and semantic categories. The contribution of this dissertation is such a streamlined overall process from the sensor data capturing over the enhancement of the dataset to the access to specific subsets. The main focus lies on the access to sub-datasets based on the spatial, temporal and semantic modalities of the stored sensor data. In general, the contribution is not restricted to a specific use case domain, yet this dissertation has use cases in agricultural robotics. Thus, use cases from the agricultural robotics domain are used to demonstrate how the spatio-temporal-semantic data access enables efficient data enhancement and analysis. We claim that storing sensor data in an HDF5 file is faster than the well-known rosbag approach. We prove this with experiments comparing the data storing rate of both approaches. Additionally, we claim that the access to sub-datasets based on the three modalities at once is novel and that no other solution to this problem currently exists. We back this claim by an extensive comparison to existing solutions.