Using Context to Identify Difficult Driving Situations in Unstructured Environments

Authors:
Kevin R. Dixon;Justin D. Basilico;Chris Forsythe;Wilhelm E. Kincses
Affiliations:
Sandia National Laboratories, Albuquerque, USA 87185;Sandia National Laboratories, Albuquerque, USA 87185;Sandia National Laboratories, Albuquerque, USA 87185;Daimler AG Group Research, Sildelfingen, Germany 71059
Venue:
FAC '09 Proceedings of the 5th International Conference on Foundations of Augmented Cognition. Neuroergonomics and Operational Neuroscience: Held as Part of HCI International 2009
Year:
2009

Citing 9
Cited 0

Practical methods of optimization; (2nd ed.)

Practical methods of optimization; (2nd ed.)
Vision and navigation for the Carnegie-Mellon navlab

IEEE Transactions on Pattern Analysis and Machine Intelligence - Special Issue on Industrial Machine Vision and Computer Vision Technology:8MPart
Machine Learning for the Detection of Oil Spills in Satellite Radar Images

Machine Learning - Special issue on applications of machine learning and the knowledge discovery process
Time Series Segmentation for Context Recognition in Mobile Devices

ICDM '01 Proceedings of the 2001 IEEE International Conference on Data Mining
Pattern Classification (2nd Edition)

Pattern Classification (2nd Edition)
Incorporating Temporal Context with Content for Classifying Image Collections

ICPR '04 Proceedings of the Pattern Recognition, 17th International Conference on (ICPR'04) Volume 2 - Volume 02
Winning the DARPA grand challenge with an AI robot

AAAI'06 Proceedings of the 21st national conference on Artificial intelligence - Volume 1
Secure Mobility and the Autonomous Driver

IEEE Transactions on Robotics
Fast obstacle detection for urban traffic situations

IEEE Transactions on Intelligent Transportation Systems

Quantified Score

Hi-index	0.01

Visualization

Abstract

We present a context-based machine-learning approach for identifying difficult driving situations using sensor data that is readily available in commercial vehicles. The goal of this system is improve vehicle safety by alerting drivers to potentially dangerous situations. The context-based approach is a two-step learning process by first performing unsupervised learning to discover meaningful regularities, or "contexts," in the vehicle data and then performing supervised learning, mapping the current context to a measure of driving difficulty. To validate the benefit of this approach, we collected driving data from a set of experiments involving both on-road and off-road driving tasks in unstructured environments. We demonstrate that context recognition greatly improves the performance of identifying difficult driving situations and show that the driving-difficulty system achieves a human level of performance on cross-validation data.