Browse Publications Technical Papers 2017-01-0110

Robust Traffic Vehicle Lane Change Maneuver Recognition 2017-01-0110

The ability to recognize traffic vehicles’ lane change maneuver lays the foundation for predicting their long-term trajectories in real-time, which is a key component for Advanced Driver Assistance Systems (ADAS) and autonomous automobiles. Learning-based approach is powerful and efficient, such approach has been used to solve maneuver recognition problems of the ego vehicles on conventional researches. However, since the parameters and driving states of the traffic vehicles are hardly observed by exteroceptive sensors, the performance of traditional methods cannot be guaranteed. In this paper, a novel approach using multi-class probability estimates and Bayesian inference model is proposed for traffic vehicle lane change maneuver recognition. The multi-class recognition problem is first decomposed into three binary problems under error correcting output codes (ECOC) framework. With probability estimates from the three binary classifiers, multi-class probability estimates are obtained through paired team comparisons. A sequence of the multi-class probability estimates are then fed into the Bayesian inference model. The Bayesian inference model views the input as sample of a random variable, and the output of the Bayesian inference model is used for the final recognition. A data set which is collected from a real-time driving simulation platform is used for the training of the binary classifiers. Typical samples are used to evaluate the performance of the proposed approach. The experimental results have demonstrated the improvement of robustness when using the proposed approach, and the approach is able to recognize lane change maneuver of the traffic vehicle with an average prediction horizon of 1.51 seconds.


Subscribers can view annotate, and download all of SAE's content. Learn More »


Members save up to 18% off list price.
Login to see discount.
Special Offer: Download multiple Technical Papers each year? TechSelect is a cost-effective subscription option to select and download 12-100 full-text Technical Papers per year. Find more information here.