Vehicle activity segmentation from position data

Electronic vehicle guidance systems have gained much popularity over the last years. The massive use of inexpensive global positioning system receivers, combined with the rapidly increasing availability of wireless communication infrastructure, suggests that large amounts of data combining both modalities will be available in a near future. The approach presented here draws on machine learning techniques and processes logs of vehicle position data to consistently infer activities and actions carried out by one or more vehicles. A fully probabilistic activity segmentation model is introduced and specific optimization methods are applied in order to learn the model parameters in a completely unsupervised manner. Experimental results with data from large mining operations are presented to validate the new model.