Identifying Truck Stops from a Large Stream of GPS Data via a Hidden Markov Chain Model

摘要: This paper proposes and shows the application of Hidden Markov Model (HMM) to identify truck trip segments extract activity non-activity stops from large scale GPS data while accounting for spatiotemporal properties points, hence overcoming limitations existing clustering practices in freight studies. Individual trajectories different classes were extracted four years raw Australian high-performance trucks, totalling more than 71 million records. The framework presented this is easy re-apply can be transferred other datasets, without requiring secondary sources such as drivers’ logbooks infer ends. results unveil three distinct type across (stops due traffic congestion, activity, stops), their duration, providing empirical insights into diverse aspects long-haul transportation. validation showed satisfactory where modelled trajectory episodes yielded similar known trajectories, showing a 97% accuracy rate identifying stops. that percentage model incorrectly assigned any state was 3% (false negative error), instances mistakenly identified by only 1%.