Promoting excellence in mobility engineering

  1. FISITA Store
  2. Technical Papers

Classification of Driving Skills Based on Machine Learning
FISITA2010/F2010E015

Authors

Naiwala P. Chandrasiri* - Toyota InfoTechonolgy Center, Japan
Kazunari Nawa - Toyota InfoTechonolgy Center, Japan
Ichiro Kageyama - Nihon University, Japan

Abstract

Classification of driving skills/driver state is a basic issue in building driver support and infotainment systems that can be adapted to individual needs of a specific driver. In this paper, we present a machine learning approach to this problem. The concept is to learn a driver model from sensor data related to the driving environment, vehicle response and driving behavior of a driver. Once the model is built, the driving skills of a driver can be classified in a novel situation automatically. Based on a survey, we selected two driving scenarios that were difficult to drive, especially for novice drivers. They were ‘narrow straight lane’ and ‘double lane change’. For the experiment, we selected two distinctive groups of drivers; novice drivers and experienced drivers. We conducted real vehicle experiments to collect driving data on test tracks. In this paper, we focused on steering behavior and tested both steering wheel angle and its frequency coefficient as features for classification of driving skills. Machine learning algorithms; SVM (Support Vector machine) and PNN (Probabilistic Neural Network) were used for learning and testing due to their high performance abilities and wide usage. High classification rates for novice and experienced drivers were obtained. Novice drivers were slower to react, and they had larger steering wheel adjustments compared with experienced drivers.

KEYWORDS – Driver support system, Novice driver, Experienced driver, SVM, PNN

Add to basket