Transformer-based Travel Time Estimation Method for Plateau and Mountainous Environments

Abstract Current research and applications on travel time estimation primarily concentrate on urban scenarios and are unable to fulfil the necessary criteria in regions characterized by plateau and mountainous topography. In this study, a set of terrain-weather features, comprising attributes of terrain and weather, was devised to reflect the factors influencing the travel time in such settings, catering for predicating travel times in those environments. A Transformer-based travel time estimation model was developed, which can improve the estimation accuracy in paths with extensive trajectory points by leveraging the long-distance dependency capture capabilities. A meta-learning strategy was employed to enhance the model's generalization ability and ensure its applicability to accurate estimation in a range of challenging environments. The experimental results substantiated the superior performance and applicability of the proposed method and demonstrated that it is capable of meeting the travel time estimation requirements in complex environments.