Trafficability Prediction and Route Planning for Forest Machines

The objective of this project is to develop novel Machine Learning (ML) methods for predicting terrain trafficability for forest machines based on model-data fusion and to develop efficient route planning approaches based on the estimated trafficability maps with uncertainties. We hypothesize that reliable terrain trafficability predictions will be achieved by combining the multi-source heterogeneous spatiotemporal environmental open big data to in-situ measurements from forest vehicle fleet and correct complexity physical terrain models via ML methods. The second main objective of the research, automated route planning, gives the basis for the actual autonomous operational ability when combined with the local sensor information providing situational awareness. Followed by route planning formulation with constraints and boundary conditions both heuristic and maximum margin planning optimization approaches are utilized.