Nikolaos Servos
PhD Student
Robert Bosch GmbH
PhD Student
Robert Bosch GmbH
Travel time prediction for multimodal freight transports using machine learning
Predicting an accurate travel time for freight transports provides a significant value to the supply chain participants and their logistics quality. The basic requirement is a continues monitoring of freight transports, e.g., using mobile sensors. Despite the superior capabilities of Machine Learning (ML) methods to deal better with non-linear relationships, only a minority of recent publications has dealt with ML for travel time prediction in freight transports. Based on the literature, we have selected Extra Trees, AdaBoost and SVR as these methods can deal with a low volume of data and a high complexity at the same time. Using different feature combination, derived from the data, several models have been build. The models have been evaluated using real world data of multimodal container transports from Bremen, Germany, to Vance, USA and compared to historical approaches.
