Convolutional Neural Networks for Pavement Roughness Assessment Using Calibration-Free Vehicle Dynamics

Case ID:

This technology is a method of identifying road roughness (i.e., International roughness index, IRI) by collecting sensor data from a smartphone app or other computing device and applying deep learning techniques to estimate the road’s roughness index.

Monitoring of road surface conditions is a logistical challenge. Traditionally, road condition monitoring for maintenance and repair relied on driver reporting and expensive and infrequent comprehensive surveys. As a result, roadway condition data tended to be poor quality and out of date. Recent alternative solutions have made use of smartphone sensor information to automatically provide road condition information in a fast and cost-effective manner. However existing solutions suffer from either a lack of accuracy, the need for mounting extra sensors to the axle of vehicles, or calibration of the tool to provide validated baselines to compare with.

This technology leverages modern advancements in deep learning to provide high quality estimates of the roughness of a road’s surface without the need of calibration or extra hardware.


  • Municipal road maintenance
  • Smart traffic management


  • Simple smartphone app approach
  • Inexpensive
  • Accurate
  • More timely than traditional road condition surveying solutions
  • Requires no calibration / baselining
  • Requires no installation of specialized hardware
  • Any passenger vehicle can be a road roughness sensor without calibration
Patent Information:
Contact For More Information:
Tariq Ahmed
Sr Licensing Manager, College of Engineering
The University of Arizona
Lead Inventor(s):
Jong-Hyun Jeong
Hongki Jo
Gregory Ditzler