Eelis Peltola

Automated operation and functionalities of mobile machinery, including path planning and motion planning, are dependent on accurate and reliable evaluation of the traversability of the surrounding ground area. This research develops novel perception solutions for segmenting the drivable area from the sensor view of a mobile machine in harsh and unstructured environments, such as ground surfaces featuring snow, vegetation, or gravel, and off-road areas.
The output of our methods can then be used in the mapping modules of automated mobile machinery working in forests, construction sites, mines, and other rough terrain.
Our research focuses mainly on exploring deep learning methodologies for the perception sensors commonly utilized on automated mobile machinery: camera and LiDAR. Because training data from harsh environments and unstructured terrain scenarios is scarce, special focus is given to expanding the capability of drivability estimation models in scenarios where little training data is available. This is done by utilizing simulated and augmented data together with real-world data.
All models are validated in real-world scenarios and with different datasets to ensure applicability and that the research does not overfit to the training scenario. Foundation models based on recent transformer architectures are explored for the segmentation task.