Aggressive Driving Saves More Time? Multi-task Learning for Customized Travel Time Estimation

Abstract

Estimating the origin-destination travel time is a fundamental problem in many location-based services for vehicles, e.g., ride-hailing, vehicle dispatching, and route planning. Recent work has made significant progress to accuracy but they largely rely on GPS traces which are too coarse to model many personalized driving events. In this paper, we propose Customized Travel Time Estimation (CTTE) that fuses GPS traces, smartphone inertial data, and road network within a deep recurrent neural network. It constructs a link traffic database with topology representation, speed statistics, and query distribution. It also uses inertial data to estimate the arbitrary phone’s pose in car, and detects fine-grained driving events. The multi-task learning structure predicts both traffic speed at public level and customized travel time at personal level. Extensive experiments on two real-world traffic datasets from Didi Chuxing have demonstrated our effectiveness.