A Novel Heading Angle Estimation Methodology for Land Vehicles Based on Deep Learning and Enhanced Digital Map

In this paper, a novel heading angle estimation methodology for land vehicles using low-cost sensors is proposed by combining the advantages of deep learning and enhanced digital map. First, an intelligent perception model of heading-related information with hybrid structure is designed to estimate the angle difference between the vehicle driving direction and the road direction. The intelligent perception model comprises of a CNN based feature extraction model and a LS-SVM based nonlinear regression model. The extracted senior features are utilized as the input of LS-SVM to predict the angle difference between the vehicle driving direction and the road direction. Then, an enhanced digital map is established to provide the road direction through map matching. Finally, the heading angle of the vehicle is obtained by combining the angle difference and the road direction. In theory, the proposed heading angle estimation methodology is not affected by the complex urban environment and is immune to cumulative errors. To verify the feasibility and effectiveness of the proposed methodology, field experiments on different road types and with different driving maneuvers were performed. The experimental results indicate that the proposed methodology can achieve accurate heading angle estimation.