Research On Travel Time Estimation With Limited Samples

With the development of science and technology,the transportation system becomes more and more important,while it brings many traffic problems,such as traffic congestion.Traffic congestion not only affects people’s travel experience,but also puts great pressure to traffic system.Travel time estimation is very significant to alleviate this problem.Accurate travel time estimation can help people avoid congested roads to save travel time and ease traffic pressure.Existing methods mainly utilize a large number of historical trajectories to train models.However,in some scenarios,travel time estimation may face few-shot problem.This paper focuses on two scenarios with limited samples as follows.(1)En route travel time estimation(ER-TTE)aims to predict the travel time on the remaining routes.Since the traveled and remaining parts of a trip usually have some common characteristics like driving speed,it is desirable to explore these characteristics for improved performance via effective adaptation.This yet faces the severe problem of data sparsity due to the few sampled points in a traveled partial trajectory.To this end,we propose a novel adaptive meta-learning model.In particular,Since trajectories with different external information(e.g.,departure time)tend to have different characteristics,we utilize soft-clustering and derive cluster-aware initialized parameters to better transfer the shared knowledge across trajectories with similar external information.Besides,we adopt a distribution-aware approach for adaptive learning rate optimization,so as to avoid taskoverfitting which will occur when guiding the initial parameters with a fixed learning rate for tasks under imbalanced distribution.(2)Transportation-mode limited travel time estimation(TL-TTE)aims to estimate the travel time of a path in a specific transportation mode(e.g.,walking,driving).Different from traditional travel time estimation,it requires to consider the heterogeneity of transportation modes due to different moving characteristics in different modes.As a result,when applying classical travel time estimation models,sufficient data is needed for each mode to capture mode-dependent characteristics separately.While in reality,it is hard to obtain enough data in some modes,resulting in a severe data sparsity problem.A practical method to solve this problem is to leverage the mode-independent knowledge(e.g.,time of waiting for traffic lights)learned from other modes.To this end,we propose a meta-optimized method called MetaMG,which learns well-generalized initial parameters to support effective knowledge transfer across different modes.Particularly,to avoid negative transfer,we integrate a spatial-temporal memory in meta-learning to cluster trajectories according to spatial-temporal distribution similarity for enhanced knowledge transfer.Besides,a multi-granularity trajectory representation is adopted in our base model to explore more useful features in different spatial granularities while improving the robustness.