Advertisement

Teenage Driving Behavior Modeling Using Deep Learning for Driver Behavior Classification

  • Muhamad-Husaini Abu-BakarEmail author
  • Rizal Razuwan
  • Syafiq Kamal
Conference paper

Abstract

Speed transition is mainly contributed from road condition which is the change from straight to cornering path. This transition introduces a sharp non-linearity in driving speed, and this behavior is significantly changed with different driver ages. Because the non-linearity has occurred, the current linear model is inaccurate to modeling the driving speed. The aim of this paper is to model the driving speed which has a sharp transition in the road path. 100 samples of driving data from the designated track were used as an input to Deep Learning (DL) architecture. DL architecture that combines a linear model that is fitted using L1 regularization, a sequence of ReLU activation layers and develop with Keras framework associated with R studio. The designated track has a distance of 700 m with 4 corners and 4 straight paths, and the driver is among the teenager’s age in the range of 20 to 25 years old. The input for the DL is GPS coordinates and output are driving speed for each coordinate. As a result, DL model successfully developed with 4% error as compared with validation data. Even though the error increases at the (selected location) transition point with maximum value was 6%, the value is considered small for the proposed model to be accepted. As a conclusion, the model has a capability in modeling the sharp non-linearity in the road path. The model further significantly improves the driver behavior for early crash prediction.

Keywords

Deep learning Driving behavior Naturalistic driving data Machine learning 

References

  1. 1.
    Zhang, J., Wang, F.Y., Wang, K., Lin, W.H., Xu X., Chen, C.: Data-driven intelligent transportation systems: a survey. IEEE Trans. Intell. Transp. Syst. 12(4), 1624–1639 (2011).  https://doi.org/10.1109/tits.2011.2158001
  2. 2.
    Schroedl, S., Zhang, W.: Predicting driving speed using neural networks. In: Proceedings of the 2003 IEEE International Conference on Intelligent Transportation Systems, vol. 1, pp. 402–407 (2003).  https://doi.org/10.1109/itsc.2003.1251985
  3. 3.
    Zhu, X., Yuan, Y., Hu, X., Chiu, Y.C., Ma, Y.L.: A Bayesian network model for contextual versus non-contextual driving behavior assessment. Transp. Res. Part C Emerg. Technol. 81, 172–187 (2017).  https://doi.org/10.1016/j.trc.2017.05.015
  4. 4.
    Polson, N.G., Sokolov, V.O.: Deep learning for short-term traffic flow prediction. Transp. Res. Part C: Emerg. Technol. 79, 1–17 (2017). http://doi.org/10.1016/j.trc.2017.02.024
  5. 5.
    Dyrud, L., Jovancevic, A., Brown, A., Wilson, D., Ganguly, S.: Ionospheric measure-ment with GPS: receiver techniques and methods. Radio Sci. 43(6), 1–11 (2008). http://doi.org/10.1029/2007RS003770

Copyright information

© Springer Nature Singapore Pte Ltd. 2019

Authors and Affiliations

  • Muhamad-Husaini Abu-Bakar
    • 1
    Email author
  • Rizal Razuwan
    • 1
  • Syafiq Kamal
    • 1
  1. 1.System Engineering and Energy Laboratory, Section of ManufacturingUniversity Kuala Lumpur Malaysian Spanish InstituteKulimMalaysia

Personalised recommendations