Road surface condition identification based on statistical pattern recognition method

This paper proposes an effective online approach to identify road surface conditions using statistical pattern recognition techniques. This approach is designed to work under normal driving conditions, and does not require extra sensors for most vehicles produced today. In the approach, road surface conditions are first classified into a set of discrete pattern states such as dry, wet, snowy, icy, etc. An index is then defined that reflects the vehicle understeer characteristics such that it is more sensitive to the road surface conditions, but less sensitive to vehicle operations. The probability distributions of the index on each of the pre-classified road surface patterns are constructed beforehand through vehicle testing, serving as a database. The index is calculated online readily using commonly available sensor measurements, and Bayes rule is applied to recursively update the conditional probability estimate of the vehicle being on each of the pre-classified road surfaces in the database, given the calculated index. Finally the maximum-likelihood estimates for the probability of each of the pre-classified road surfaces are computed. The presented method of road surface condition identification is verified through both computer simulation and vehicle testing, which shows that the proposed approach is effective and robust in identifying the road surface conditions.


  • English

Media Info

  • Media Type: Web
  • Features: References;
  • Pagination: 6p
  • Monograph Title: 12th International IEEE Conference on Intelligent Transportation Systems (ITSC 2009)

Subject/Index Terms

Filing Info

  • Accession Number: 01574503
  • Record Type: Publication
  • ISBN: 9781424455195
  • Files: TRIS
  • Created Date: Aug 14 2015 6:09PM